DETAILED ACTION
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
2. Claims 1-20 are currently pending. Claims 1, 6-8, 13-15 and 20 have been amended. Claims 1-20 have been rejected.
Status of the Application
3. Claims 1-20 are currently pending and have been examined in this application. This communication is the first action on the merits.
Response to Amendments
4. Applicant’s amendment filed on 11/10/2025 necessitated new grounds of rejection in this office action.
Continued Examination under 37 CFR 1.114
5. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 01/22/2026 has been entered.
Foreign Priority
6. The Examiner has noted the Applicants claiming Priority from Foreign Application GR20230100676 filed on 08/16/2023. Receipt is acknowledged of papers submitted under 35 U.S.C. § 119(a)-(d), which papers have been placed of record in the file. Examiner Note: Applicant submits Certified Copy of Foreign Priority Application on 01/28/2026 which is recognized by the Examiner as acceptable for consideration of Foreign Priority Date purposes for this application.
Response to Arguments
7. Applicant’s arguments, see pages 12-14 of 18 filed on 11/10/2025, with respect to the 35 U.S.C. § 101 Claim Rejections for Claims 1-20 have been fully considered and are found to be persuasive.
Therefore, the 35 U.S.C. § 101 Claim Rejections for Claims 1-20 have been withdrawn. See the 35 U.S.C. § 101 Subject Matter Eligibility Analysis Section below explaining why Claims 1-20 are deemed patent eligible over 35 U.S.C. § 101.
8. Applicant’s arguments, see pages 14-17 of 18 filed on 11/10/2025, with respect to the 35 U.S.C. § 103 Claim Rejections for Claims 1-20 have been fully considered and are found to be not persuasive. Applicant’s arguments with respect to Claims 1-20 have been considered, but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
35 U.S.C. § 101 Subject Matter Eligibility Analysis
9. 35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
10. Step 1: Claims 1-20 are focused to a statutory category namely, a “method” or a “process” (Claims 1-7), a “system” or an “apparatus” (Claims 8-14) and a “non-transitory machine-readable medium” or “article of manufacture” (Claims 15-20).
Step 2A Prong One: Step 2a: Prong 1 – Is the Claim Directed to a Judicial Exception?
For Independent Claims 1, 8 and 15, the steps include training models, receiving data, partitioning/classifying text, generating a time series, and updating it. This could be viewed as a "method of organizing human activity" (managing well-being) or "mental processes" (evaluating text). However, these claims are not merely abstract because they are not performing these tasks in the human mind, nor is it a basic mathematical formula applied to a generic field. It is a specific technical solution using AI. Training & Retraining: Training algorithms and adjusting weights (e.g., retraining) are technical processes, not abstract ideas, according to recent PTAB decisions (e.g., Ex parte Desjardins), which noted that dynamic, continuous learning techniques are not merely "mental processes" but rather technical improvements to AI model performance. Text Processing: The partitioning and classification, when executed specifically by "one or more processors," represent a concrete, automated process.
Conclusion (Step 2a - Prong 1): Independent Claims 1, 8 and 15 are likely not directed to an abstract idea, but even if it were, it passes Prong 2.
Therefore, at step 2a prong 1, Yes, Claims 1-20 do not recite an abstract idea under step 2a prong 1.
Step 2a: Prong 2 – Does the Claim Integrate the Exception into a Practical Application?
Even assuming arguendo that the Independent Claims 1, 8 and 15 do not recite an abstract idea, these claims integrate any potential abstract idea into a practical application, making it patent-eligible under the 2024 USPTO guidelines with respect to Step 2A Prong Two of the eligibility inquiry (as explained in MPEP § 2106.04(d)). For Independent Claims 1, 8 and 15, these steps of the claimed process, particularly the real-time retraining and weighting adjustment (retraining, adjusting one or more weights), provides a concrete technical improvement to machine learning, and the entire system constitutes a specific, practical application of AI in a digital, networked environment via MPEP § 2106.04(a).
For example; Independent Claim 1 is broken down according to the following steps:
Step 1: "Receiving... from one or more applications of a digital workspace": This is a specific technological implementation, not abstract data gathering.
Step 2: "Partitioning... into text units that are each a minimal body of text...": This is a technical, automated improvement in data segmentation (NLU) rather than a mere "mental" classification.
Step 3: "Generating a time series distribution... in a continuous and opaque manner": This is a specific, non-routine application of data visualization technology to provide real-time, actionable insights.
Step 4: "Dynamically updating... in real-time...": This adds a specific, non-routine, technical step that improves the functionality of the system, addressing the "continuous monitoring" requirement.
Step 5: "Retraining... adjusting one or more weights...": This specifically integrates the AI model into an improvement of the technical system itself, mitigating issues like accuracy decay over time, which is a recognized technical improvement. The process of dynamically retraining the AI models using updated, real-time data to adjust weights based on a comparison of previous outputs and new outcomes is a specific technical, iterative, and "non-routine" step. "Continuous and Opaque Manner": Designing a system that processes data while managing user privacy ("opaque manner") and reducing bias (a technical challenge) demonstrates that the invention provides specific, non-conventional technical solutions to technical problems. Integration with Digital Workspace: These claims are not just doing math in a vacuum; it is applying AI to a "digital workspace of an organization," a specific technological environment.
Therefore, Independent Claims 1, 8 and 15, the combination of real-time text segmentation, adaptive AI retraining/reweighting, and, specifically, the continuous update loop constitutes reciting additional elements that integrate the judicial exception into a practical application by (1) improvements to the functioning of a computer, or to any other technology or technical field (see MPEP § 2106.05 (a)) or (2) applying or using the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claims as a whole is more than a drafting effort designed to monopolize the judicial exception (see MPEP § 2106.05 (e)).
Conclusion (Step 2a - Prong 2): Yes, the steps for Independent Claims 1, 8 and 15 provide a specific, non-routine, technical application of AI to the problem of organizational well-being. Thus, Claims 1-20 are patent eligible under 35 U.S.C. § 101 step 2a prong 2 of the 35 U.S.C. § 101 analysis.
Claim Rejections - 35 USC § 103
11. This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
12. In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
13. The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
14. Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over US PG Pub (US 2024/0054430 A1) hereinafter Maikhuri, et. al., in view of US PG Pub (US 2024/0184989 A1) hereinafter Mann, et. al., in view of U.S. Patent # (US 10,860,805 B1) hereinafter Coppersmith, et. al., in view of US PG Pub (US 2020/0090067 A1) hereinafter Anders, and in further view of US PG Pub (US 2024/0403794 A1) hereinafter Shook, et. al.
Regarding Independent Claim 1, Maikhuri method for using artificial intelligence models to generate a time series distribution of text units teaches the following:
- training (see at least Maikhuri: Fig. 7 & Fig. 10 & ¶ [0119]. Maikhuri notes training content 136, and the associated analyses with these modules, to help make complex decisions for the user. The PARE 105 is configured to perform personal analysis on the user 113 a (via the user interactions 114 that are recorded and analyzed) and make one or more recommendations (e.g., in the form of first information 130 that includes, but is not limited to, assessments, user reports, recommendations, and/or feedback), with goals of helping a user to understand, analyze, and improve personal and professional user behavior and interactions, to meet personal effectiveness goals.), by one or more processors (see at least Maikhuri: Fig. 14 & ¶ [0202]. Maikhuri teaches a processor CPU 1402 shown in Fig. 14.) using a training algorithm (see at least Maikhuri: ¶ [0169] & ¶ [0176-0178]. Maikhuri notes that emotion Classification 812 is the classification algorithm to classify emotions based on the extracted features. The classification has various methods, which classify the images into multiple classes. See also Maikhuri at ¶ [0111]: “This intelligent and specifically trained IPAE 102 (and optional associated robotic automation is) configured to understand the employees' context, role, domain, intent, and communication/task.” See also Maikhuri at ¶ [0176-0178]: “This model uses historical training data containing multi-dimension data points to train the model. Once the model is fully trained, the conversation's state (intent, sentiment, context) is passed to predict the following best action. The algorithm Random Forest uses a large group of complex decision trees and can provide classification predictions with a high degree of accuracy on any size of data. This engine algorithm will predict the recommended virtual assistant with the accuracy or likelihood percentage.”.), a first artificial intelligence (AI) model to partition text content into text units (see at least Maikhuri: Fig. 1A & ¶ [0116] & ¶ [0128] & ¶ [0162]. Maikhuri teaches that the text/natural language processing module 104 is configured for entity recognition and content classification, as well as intent and sentiment analysis. The text/natural language processing module 104 also is configured to cooperate with outputs of the voice/audio analysis module 106 and video analysis module 108 to derive and/or determine intent and sentiment from that content, as well as to assist in performing content classification. See also Maikhuri at ¶ [0027]: “The first analysis configured to analyze the set of raw data records for at least one of sentiments, emotions, and intent; performing a second analysis on the set of raw data records, the second analysis configured to segment the set of raw data records.” See also Maikhuri at ¶ [0128]: The second analysis may use automatic text segmentation 166, which is configured to breaks up text into topically-consistent segments. Text segmentation, as is known, is related to natural language processing, document classification, and information retrieval. In block 165, using text segmentation, specific features can be extracted and segmented in the data, such as determinations that, within the raw user data 144, a specific project, context, individual, etc., is mentioned, or that a specific string of words are related and together represent an actionable action or statement. See also Maikhuri at ¶ [0162]: “The artificial intelligence (AI) 422 further processes this information.”) and a second AI model (see at least Maikhuri: ¶ [0092] & ¶ [0115] & ¶ [0126]. Maikhuri teaches that the analysis of the intelligent processing engine 101 uses Natural Language Processing (NLP) to understand the communication context, intent, sentiment, etc., and sends these details to the personal knowledge/domain expertise repository. A part of this functionality is the IPAE's ability to understand and express itself in natural language, creating an intuitive way for the employee/person to perform tasks.) to classify the text units into respective well-being-related categories (see at least Maikhuri: ¶ [0126] & ¶ [0164]. Maikhuri teaches that sentiment analysis a process of using automated processes and/or machine learning, such as natural language processing (NLP), text analysis, and statistics to analyze the sentiment in one or more of string of words (such as an email, a social media post, a statement, a text, etc.). Sentiment analysis includes techniques and methods for understanding emotions by use of software. Natural language processing, statistics, and text analysis are used as part of sentiment analysis to extract, and identify the sentiment of words (e.g., to determine if words may be positive, negative, neutral, and/or whether there are additional emotions that can be inferred, such as anger, confusion, enthusiasm, humor, etc.). See also Maikhuri at ¶ [0147]: “A general-purpose entity can correspond to a storage location having an identifier that may be common to many users, such as “to-do list,” “summary of meetings,” “feedback from boss,” “overdue tasks.” When raw user data is analyzed, the intelligent processing engine 101 may determine that, based on the processed and interpreted content, a given raw user data record may fit into one of the predetermined entity categories.” See also Maikhuri at ¶ [0164]: Sentiment analysis is done through text data. Audio data also is processed to help detect a person's emotions just by their voice which will help to know and interpret that person's actions and/or their behavior. Neural network techniques such as multilayer perceptron (MLP) and long short-term memory (LSTM) are less advantageous, so techniques such as convolutional neural network (CNN) are used to classify in the problem/situation where different emotions need to be categorized.) based on training data including inputs and known outcomes (see at least Maikhuri: ¶ [0157] & ¶ [0177] & Fig. 7 & Fig. 10. Maikhuri notes that the IPAE 102, after analyzing the communication, can dynamically create first information 130 such as an assessment summary that includes specific recommendations, tasks, or brief summaries 202 of all messages and chats that day, so that the user 113 a does not have to go back and review multiple threads or chats for final outcome and actions. See also Maikhuri at ¶ [0112-0114]: User interactions inputs, and related user interaction data, including inputs input as speech/text and audio and sends/converts output as speech/text/audio, as applicable. For example, the communications gateway 103 can receive information in text, video, and/or audio format. See also Maikhuri at ¶ [0129]: The IPAE 102 classifies received raw user data 144 (e.g., set of user data, user interactions, inputs, and optional related user interaction data) to organize by sentiment, emotion, intent, domain, project, environment, task, interaction group, etc.)
Maikhuri method for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Mann in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- receiving, by the one or more processors (see at least Mann: ¶ [0295-0297].) and from one or more applications of a digital workspace of an organization (see at least Mann: ¶ [0032] & ¶ [0434] & ¶ [0603]. Mann notes tracking electronic connections between a plurality of entities in an electronic workspace; tracking characteristics of the electronic connections between the plurality of entities in the electronic workspace. See also Mann at ¶ [0434]: Workspace templates icon 2614 may enable users to package boards and dashboards as a unified solution. See also Mann at ¶ [0603]: One or more automations in an automation package may be customized for a profession, a vocation, an industry, a technology, an occupation, a business, or other entities having collaborative workspaces. One or more automations in an automation package may also be customized based on specific use cases for certain tasks, such as tracking project progress, enabling communications between remote individuals, or managing files between teams. See also Mann at ¶ [0636]: A collaborator may use a primary application as a main working environment, while functionality of third-party applications may be integrated into the primary application. See also Mann at Fig. 141 of the digital workspaces.), the text content generated by users of the organization (see at least Mann: ¶ [0549] & ¶ [1396]. Mann notes characterizing text messages may include recording key words from the text messages. The system may analyze the logged chat messages from a communication and determine key words or phrases spoken by each individual or all individuals during the video or audio communication. This may enable users to track action items at the conclusion of the communication. The system may enable participants to manually mark key words from the communication or text messages for recordation so that the participants may later refer to the key word. See also Mann at ¶ [1396]: Mann notes that semantic analysis involves a computer process for drawing meaning from text. It may involve identifying relationships between individual words in a particular context within sentences, paragraphs, or whole documents by electronically analyzing grammatical structure and identifying relationships between particular words in a particular context. After semantic analysis is performed on first data in a first board and second data in a second board, the at least one processor can compare the results to determine a similarity.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri method for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: receiving, by the one or more processors and from one or more applications of a digital workspace of an organization, the text content generated by users of the organization, and in view of Mann, whereby the relationship or connection may be established by linking the automation and the table, or by assigning a common code, address, or other designation to the automation and the table. One or more automations in an automation package may be customized for a profession, a vocation, an industry, a technology, an occupation, a business, or other entities having collaborative workspaces. One or more automations in an automation package may also be customized based on specific use cases for certain tasks, such as tracking project progress, enabling communications between remote individuals, or managing files between teams (see at least Mann: ¶ [0603]).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Mann, the results of the combination were predictable.
Maikhuri / Mann method for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Coppersmith in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- partitioning, by the one or more processors (see at least Coppersmith: Fig. 1a & Col. 16, Lns. 54-60. Coppersmith noting a processor shown in Fig. 1a.) and using the first AI model (see at least Coppersmith: Col. 1, Lns. 63-67 & Col. 2, Lns. 65-67 & Col. 3, Lns. 1-2. Coppersmith notes that Figs. 1a, 1b and 1C, computer system 100 may use NLP and other artificial intelligence techniques to evaluate Slack messages 110 and other free-form text 112, other sensor monitoring 114 of team members and their environment, behavior and communication of workers or team members of a team, and surveys to assess emotional content, sentiment, and psychological well-being of the team members and of the team as a whole.), the text content into the text units (see at least Coppersmith: Col. 3, Lns. 1-7 & Col. 6, Lns. 50-54 & Col. 7, Lns. 46-60. Coppersmith notes that emotion analysis 132 may analyze for emotions such as joy, sadness, surprise, anger and fear and a residual category for messages that express no emotion. See also Coppersmith at Col. 6, Lns. 50-54. Individual data streams such as Slack messages, text messages, or intra-team messages may have captured a large-enough spectrum of team communications to allow system 100 to show mental health, well-being, or stress state.) that are each a minimal body of text containing well-being-related information (see at least Coppersmith: Figs. 6B-6C & Col. 14, Lns. 44-50 & Col. 14, Lns. 60-64 & Col. 15, Lns. 1-11. Coppersmith notes that messages may be grouped by the emotion and/or sentiment. For example, if there are eight categories of emotions (seven emotions plus no emotion) and three sentiments (positive, negative, and neutral), messages may be grouped into twenty-four categories, and messages may be grouped by histogramming message counts into those twenty-four categories. In each time interval, system 100 may collect a count of the total number of messages exchanged in the group. Data may reflect some time interval, such as six months (for example, 60 three-day periods). The scores may be summed to the margins—for example, for a three-day period, the twenty-four cells may be summed into eight emotion margin cells, and three margin sentiment cells. In step 662, as a further data normalization, system 100 may compute a mean over the entire time interval, and then normalize each margin cell to a standard deviation measure. For example, if the 60 cells have an average number of 20 messages per day showing “joy” emotion, with a standard deviation of 3 counts, then cells with 16, 20, and 22 counts would be normalized to “−1.33, 0, 0.67” for further processing.);
- classifying (see at least Coppersmith: Fig. 1 denoting emotional classifiers 132 and sentiment classifiers 134.), by the one or more processors (see at least Coppersmith: Fig. 1a & Col. 16, Lns. 54-60. Coppersmith noting a processor shown in Fig. 1a.) and using the second AI model (see at least Coppersmith: Col. 1, Lns. 63-67 & Col. 2, Lns. 65-67 & Col. 3, Lns. 1-2. Coppersmith notes that Figs. 1a, 1b and 1C, computer system 100 may use NLP and other artificial intelligence techniques to evaluate Slack messages 110 and other free-form text 112, other sensor monitoring 114 of team members and their environment, behavior and communication of workers or team members of a team, and surveys to assess emotional content, sentiment, and psychological well-being of the team members and of the team as a whole.), the text units into the respective well-being-related categories (see at least Coppersmith: Col. 3, Lns. 1-7 & Col. 6, Lns. 50-54 & Col. 7, Lns. 46-60. Coppersmith notes that emotion analysis 132 may analyze for emotions such as joy, sadness, surprise, anger and fear and a residual category for messages that express no emotion. See also Coppersmith at Col. 6, Lns. 50-54. Individual data streams such as Slack messages, text messages, or intra-team messages may have captured a large-enough spectrum of team communications to allow system 100 to show mental health, well-being, or stress state.)
- generating, by the one or more processors (see at least Coppersmith: Fig. 1a & Col. 16, Lns. 54-60. Coppersmith noting a processor shown in Fig. 1a.), a time series distribution of the text units among the well-being-related categories based on classifying the text units into the respective well-being-related categories in a continuous and opaque manner that reduces bias (see at least Coppersmith: Figs. 2A – 2G & Col. 5, Lns. 20-44 & Col. 12, Lns. 51-60. Coppersmith teaches that the history of interventions will be recorded alongside the time series data and we will analyze the effect that, e.g., deliberately increasing cross-clique communication and collaboration has on the team's multiplex graph metrics and resulting measured CAL. In this way the system is closed-loop; experimental variations and perturnations may be conducted in as controlled a manner as possible, and effects of those experimental interventions measured to determine viability of future interventions to decrease CAL. See also Coppersmith at Col. 5, Lns. 20-44: “Software 100 may use persistent and passive monitoring of team communications 112, environmental monitoring 114, etc. When monitoring is passive, team members need not act to “turn it on,” which allows for the data collection to be more persistent, reducing sampling error and bias (e.g., participation bias could favor participation from those who have free time to do so, while those team members under higher allostatic load will not as easily make the time).” See also Coppersmith at Figs. 2A-2G noting ““time series distributions” of text units among well-being-related categories.” See also Coppersmith at Col. 3, Lns. 26-36: “The emotional classifiers 132 and sentiment classifiers 134 discerned from text and sensors may include suicide risk, mental illness, joy, fear, sadness, anger, or frustration among team members, team effectiveness, language between two team members (e.g., mentor-mentee, cooperative, consultative, joint problem solving, helpful disagreement, sharp, or condescending), and CAL. These analyses may provide an experimental testbed to evaluate emotional state of individuals and of the team.” See also Coppersmith at Fig. 3C and Fig. 6C as well.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann method for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: partitioning, by the one or more processors and using the first AI model, the text content into the text units that are each a minimal body of text containing well-being information & classifying, by the one or more processors and using the second AI model, the text units into the respective well-being-related categories & generating, by the one or more processors, a time series distribution of the text units among the well-being-related categories based on classifying the text units into the respective well-being-related categories in a continuous and opaque manner that reduces bias, and in further view of Coppersmith, whereby artificial intelligence techniques are used in Coppersmith to evaluate Slack messages, other free-form text, other sensor monitoring of team members and their environment, behavior, and communication of workers or team members of a team to assess emotional content, sentiment, and psychological well-being of the team’s members and of the team as a whole (see at least Coppersmith: Col. 2, Lns. 65-67 & Col. 3, Lns. 1-7)
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Coppersmith, the results of the combination were predictable.
Maikhuri / Mann / Coppersmith method for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Anders in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- dynamically updating (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. See Anders at ¶ [0025]: “By use of the emotion time series model 160, the emotion service engine 130 dynamically ascertains the state emotion of the user 105 at a certain time”.), by the one or more processors (see at least Anders: Fig. 8 & ¶ [0102] & ¶ [0104-0105]. Anders notes one or more processors 16 shown in the computer system of Fig. 8.) and in real-time, the time series distribution based on additional text content received after initial generation of the time series distribution (see at least Anders: ¶ [0039-0040] & Fig. 5 & Fig. 7. Anders notes that the emotion service engine 130 can also modify the calendar of the user 105 by adding additional alerts for an existing schedule or generating a new event on the calendar, with a notice “Don't forget the office outing is this Friday! Two more days to go!” or “The quarterly report is due in two weeks. Please have it ready by next Wednesday for a review before the presentation!”. Anders at ¶ [0040] teaches that the emotion service engine 130 also obtains updates on the emotion time series data 113 from the data collection devices 107. The emotion service engine 130 trains the emotion time series model 160, by machine learning with the updated inputs and/or the feedback 199 from the user 105, in order to improve accuracy of the emotion time series model 160 with the time progression of the baseline as well as the predictions.);
- causing, by the one or more processors (see at least Anders: Fig. 8 & ¶ [0102] & ¶ [0104-0105]. Anders notes one or more processors 16 shown in the computer system of Fig. 8.), a user interface to be displayed on a user device (see at least Anders: Fig. 8 & ¶ [0110]. Anders notes that in addition to or in place of having external devices 14 and the display 24, which can be configured to provide user interface functionality, computing node 10 can include another display 25 connected to bus 18.), the user interface (see at least Anders: Fig. 8 & ¶ [0110].) including the dynamically updated (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. See Anders at ¶ [0025]: “By use of the emotion time series model 160, the emotion service engine 130 dynamically ascertains the state emotion of the user 105 at a certain time”.) time series distribution of the text units among the well-being-related categories for the organization (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. Anders teaches that the emotion service engine 130 also obtains updates on the emotion time series data 113 from the data collection devices 107. The emotion service engine 130 trains the emotion time series model 160, by machine learning with the updated inputs and/or the feedback 199 from the user 105, in order to improve accuracy of the emotion time series model 160 with the time progression of the baseline as well as the predictions. See also Anders at ¶ [0044]: The emotion service engine 130 processes the collected text contents and classifies the text contents into certain categories indicative of states of emotion, by use of the external natural language understanding tools 170, which includes topic modeling tools. All emotion time series data 113 are respectively associated with a unique time stamp that records the time upon which the emotion time series data 113 had been observed/collected on the user 105. See also Anders at Fig. 5 noting an “Emotion Time Series” 500 and “Basic Emotion Time Series: Joy 510”. See also Anders at Fig. 7 & ¶ [0078]: “The application-emotion time series graph 700 in a first area 701 is for a chat application, which represents a decreasing emotion score over time for the duration between “0:00:00” and “0:18:00”, which indicates that using the chat application program has a negative effect on the emotion score.”).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith method for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: dynamically updating, by the one or more processors and in real-time, the time series distribution based on additional text content received after initial generation of the time series distribution & causing, by the one or more processors, a user interface to be displayed on a user device, the user interface including the dynamically updated time series distribution of the text units among the well-being-related categories for the organization, and in further view of Anders, whereby the emotion service engine also obtains updates on the emotion time series data. The emotion service engine trains the emotion time series model, by machine learning with the updated inputs and/or the feedback from the user, in order to improve accuracy of the emotion time series model with the time progression of the baseline as well as the predictions. The emotion service engine repeats through when preconfigured conditions for re-processing are met, including, but not limited to, cumulating a predefined amount of input data, obtaining a new instance of the target emotion-time value pair (see at least Anders: ¶ [0040].)
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Anders, the results of the combination were predictable.
Maikhuri / Mann / Coppersmith / Anders method for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Shook in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- retraining (see at least Shook: ¶ [0151] & Fig. 4. Shook teaches that data collected at the data collected 401 can be used to iteratively retrain the model to accurately predict engagement scores.), by the one or more processors (see at least Shook: Fig. 1 & ¶ [0175-0176].) using the training algorithm (see at least Shook: Figs. 6-10).), the first AI model and/or the second AI model (see at least Shook: Figs. 3-4 & ¶ [0105-0106]. Shook teaches creating a workflow pipeline that uses the combination of above-mentioned data manipulation techniques, that includes identifying inconsistencies in the data, leveraging NLP to correctly parse unstructured data, and using data normalization techniques to standardize the data. Additionally, automated validation checks are be used to ensure high quality of data used for pattern recognition and insights. Natural Language Processing/Understanding (NLP/NLU) techniques are be used to identify understand the semantics on common ground.) using updated training data based on the additional text content (see at least Shook: ¶ [0134-0135] & ¶ [0143] & ¶ [0148] & Figs. 3-4. Shook notes that data Augmentation 445 can be used to generate additional data and diversify the dataset. For example, if there's bias towards positive appraisal feedback, synthetic data points representing negative feedback can be generated to balance the distribution and balance the distribution of occurrence. Data model of the entity data can be defined and re-fined to collect relevant data for the entities that can be used for training the ML model. See also Shook at ¶ [0143]: At 470, the model is tuned for accuracy by implementing calibration and de-biasing techniques that support improved accuracy of the prediction of the engagement scores. Calibration techniques can be used to adjust the model's predictions to align with ground truth outcomes and reduce bias. Such calibration techniques can involve comparing predicted probabilities to observed outcomes and performing adjustments to the model as necessary. See also Shook at Fig. 3 noting “training data 340”.), and adjusting one or more weights of the first AI model and/or the second AI model (see at least Shook: ¶ [0143] & ¶ [0145-0147]. Shook teaches executing the ML model to forecast the set of engagement index score for the set of entities based on the data provided for the set of entities comprises: defining weights for contribution of the parameters to be used for the execution, wherein a weight adjusts a contribution of a parameter when evaluated by the ML model. The weights assigned to different influencing factors as parameters of the ML model can vary based on at least one of organizational priorities, the specific goals, and/or individual employee needs. See also Shook at ¶ [0147]: “Weights assigned to parameters of the ML model can be adjusted to account for such identified bias in the provided data for the ML model execution. By considering a combination of influencing factors on entity level, adjusting weights as needed based on identified bias in provided data for executing the ML model, a centralized source platform that runs the ML model or another system that is configured to execute the ML model can generate tailored-made action plans that are both personalized and equitable for all employees”.) based on a comparison between previous outputs and updated known outcomes (see at least Shook: ¶ [0103] & ¶ [0137-0138] & ¶ [0141-0143]. Shook notes that the obtained training data is data that is enhanced based on the techniques implemented at the data enhancement component 402. The trained model can include defined parameters as previously discussed for which data is collected from the data collector 401. See also Shook at ¶ [0103]: When the training data 340 is generated, considerations for identifying and mitigation hidden bias can be made so that different data augmentation techniques can be performed to modify the collected entity data 320 to remove bias and thus when used for training the ML model 380 to provide a more accurate model that is de-biased compared to other models that can rely on biased data. See also Shook at ¶ [0143]: At 470, the model is tuned for accuracy by implementing calibration and de-biasing techniques that support improved accuracy of the prediction of the engagement scores. Calibration techniques can be used to adjust the model's predictions to align with ground truth outcomes and reduce bias. Such calibration techniques can involve comparing predicted probabilities to observed outcomes and performing adjustments to the model as necessary. See also Shook at ¶ [0148]: Association techniques can be applied that rely on actions that are available to be performed for entities that are determined to be influences by a given influence factor so that their engagement score, when recomputed after the actions are implemented can be modified to show improved engagement compared to the initially computed engagement score.);
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith / Anders method for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: retraining, by one or more processors using the training algorithm, the first AI model and/or the second AI model using updated training data based on the additional text content, and adjusting one or more weights of the first AI model and/or the second AI model based on a comparison between pervious outputs and updated known outcomes, and in further view of Shook, by using advanced algorithms and ML techniques, personalized and curated plans to enhance engagement index scores of employees across an organization can be provided. By proactively addressing specific areas for improvement and aligning them with employee preferences, organizations can foster an engaged and motivated workforce, leading to higher overall organization-wide engagement levels (see at least Shook: ¶ [0102].).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Shook, the results of the combination were predictable.
Regarding Independent Claim 8, Maikhuri system for using artificial intelligence models to generate a time series distribution of text units teaches the following:
- a memory configured to store instructions (see at least Maikhuri: ¶ [0202] & Fig. 14.)
- one or more processors configured to execute the instructions to perform operations (see at least Maikhuri: ¶ [0202] & Fig. 14.) comprising:
- training (see at least Maikhuri: Fig. 7 & Fig. 10 & ¶ [0119]. Maikhuri notes training content 136, and the associated analyses with these modules, to help make complex decisions for the user. The PARE 105 is configured to perform personal analysis on the user 113 a (via the user interactions 114 that are recorded and analyzed) and make one or more recommendations (e.g., in the form of first information 130 that includes, but is not limited to, assessments, user reports, recommendations, and/or feedback), with goals of helping a user to understand, analyze, and improve personal and professional user behavior and interactions, to meet personal effectiveness goals.), using a training algorithm (see at least Maikhuri: ¶ [0169] & ¶ [0176-0178]. Maikhuri notes that emotion Classification 812 is the classification algorithm to classify emotions based on the extracted features. The classification has various methods, which classify the images into multiple classes. See also Maikhuri at ¶ [0111]: “This intelligent and specifically trained IPAE 102 (and optional associated robotic automation is) configured to understand the employees' context, role, domain, intent, and communication/task.” See also Maikhuri at ¶ [0176-0178]: “This model uses historical training data containing multi-dimension data points to train the model. Once the model is fully trained, the conversation's state (intent, sentiment, context) is passed to predict the following best action. The algorithm Random Forest uses a large group of complex decision trees and can provide classification predictions with a high degree of accuracy on any size of data. This engine algorithm will predict the recommended virtual assistant with the accuracy or likelihood percentage.”.), a first artificial intelligence (AI) model to partition text content into text units (see at least Maikhuri: Fig. 1A & ¶ [0116] & ¶ [0128] & ¶ [0162]. Maikhuri teaches that the text/natural language processing module 104 is configured for entity recognition and content classification, as well as intent and sentiment analysis. The text/natural language processing module 104 also is configured to cooperate with outputs of the voice/audio analysis module 106 and video analysis module 108 to derive and/or determine intent and sentiment from that content, as well as to assist in performing content classification. See also Maikhuri at ¶ [0027]: “The first analysis configured to analyze the set of raw data records for at least one of sentiments, emotions, and intent; performing a second analysis on the set of raw data records, the second analysis configured to segment the set of raw data records.” See also Maikhuri at ¶ [0128]: The second analysis may use automatic text segmentation 166, which is configured to breaks up text into topically-consistent segments. Text segmentation, as is known, is related to natural language processing, document classification, and information retrieval. In block 165, using text segmentation, specific features can be extracted and segmented in the data, such as determinations that, within the raw user data 144, a specific project, context, individual, etc., is mentioned, or that a specific string of words are related and together represent an actionable action or statement. See also Maikhuri at ¶ [0162]: “The artificial intelligence (AI) 422 further processes this information.”) and a second AI model (see at least Maikhuri: ¶ [0092] & ¶ [0115] & ¶ [0126]. Maikhuri teaches that the analysis of the intelligent processing engine 101 uses Natural Language Processing (NLP) to understand the communication context, intent, sentiment, etc., and sends these details to the personal knowledge/domain expertise repository. A part of this functionality is the IPAE's ability to understand and express itself in natural language, creating an intuitive way for the employee/person to perform tasks.) to classify the text units into respective well-being-related categories (see at least Maikhuri: ¶ [0126] & ¶ [0164]. Maikhuri teaches that sentiment analysis a process of using automated processes and/or machine learning, such as natural language processing (NLP), text analysis, and statistics to analyze the sentiment in one or more of string of words (such as an email, a social media post, a statement, a text, etc.). Sentiment analysis includes techniques and methods for understanding emotions by use of software. Natural language processing, statistics, and text analysis are used as part of sentiment analysis to extract, and identify the sentiment of words (e.g., to determine if words may be positive, negative, neutral, and/or whether there are additional emotions that can be inferred, such as anger, confusion, enthusiasm, humor, etc.). See also Maikhuri at ¶ [0147]: “A general-purpose entity can correspond to a storage location having an identifier that may be common to many users, such as “to-do list,” “summary of meetings,” “feedback from boss,” “overdue tasks.” When raw user data is analyzed, the intelligent processing engine 101 may determine that, based on the processed and interpreted content, a given raw user data record may fit into one of the predetermined entity categories.” See also Maikhuri at ¶ [0164]: Sentiment analysis is done through text data. Audio data also is processed to help detect a person's emotions just by their voice which will help to know and interpret that person's actions and/or their behavior. Neural network techniques such as multilayer perceptron (MLP) and long short-term memory (LSTM) are less advantageous, so techniques such as convolutional neural network (CNN) are used to classify in the problem/situation where different emotions need to be categorized.) based on training data including inputs and known outcomes (see at least Maikhuri: ¶ [0157] & ¶ [0177] & Fig. 7 & Fig. 10. Maikhuri notes that the IPAE 102, after analyzing the communication, can dynamically create first information 130 such as an assessment summary that includes specific recommendations, tasks, or brief summaries 202 of all messages and chats that day, so that the user 113 a does not have to go back and review multiple threads or chats for final outcome and actions. See also Maikhuri at ¶ [0112-0114]: User interactions inputs, and related user interaction data, including inputs input as speech/text and audio and sends/converts output as speech/text/audio, as applicable. For example, the communications gateway 103 can receive information in text, video, and/or audio format. See also Maikhuri at ¶ [0129]: The IPAE 102 classifies received raw user data 144 (e.g., set of user data, user interactions, inputs, and optional related user interaction data) to organize by sentiment, emotion, intent, domain, project, environment, task, interaction group, etc.)
Maikhuri system for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Mann in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- receiving, from one or more applications of a digital workspace of an organization (see at least Mann: ¶ [0032] & ¶ [0434] & ¶ [0603]. Mann notes tracking electronic connections between a plurality of entities in an electronic workspace; tracking characteristics of the electronic connections between the plurality of entities in the electronic workspace. See also Mann at ¶ [0434]: Workspace templates icon 2614 may enable users to package boards and dashboards as a unified solution. See also Mann at ¶ [0603]: One or more automations in an automation package may be customized for a profession, a vocation, an industry, a technology, an occupation, a business, or other entities having collaborative workspaces. One or more automations in an automation package may also be customized based on specific use cases for certain tasks, such as tracking project progress, enabling communications between remote individuals, or managing files between teams. See also Mann at ¶ [0636]: A collaborator may use a primary application as a main working environment, while functionality of third-party applications may be integrated into the primary application. See also Mann at Fig. 141 of the digital workspaces.), the text content generated by users of the organization (see at least Mann: ¶ [0549] & ¶ [1396]. Mann notes characterizing text messages may include recording key words from the text messages. The system may analyze the logged chat messages from a communication and determine key words or phrases spoken by each individual or all individuals during the video or audio communication. This may enable users to track action items at the conclusion of the communication. The system may enable participants to manually mark key words from the communication or text messages for recordation so that the participants may later refer to the key word. See also Mann at ¶ [1396]: Mann notes that semantic analysis involves a computer process for drawing meaning from text. It may involve identifying relationships between individual words in a particular context within sentences, paragraphs, or whole documents by electronically analyzing grammatical structure and identifying relationships between particular words in a particular context. After semantic analysis is performed on first data in a first board and second data in a second board, the at least one processor can compare the results to determine a similarity.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri system for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: receiving, from one or more applications of a digital workspace of an organization, the text content generated by users of the organization, and in view of Mann, whereby the relationship or connection may be established by linking the automation and the table, or by assigning a common code, address, or other designation to the automation and the table. One or more automations in an automation package may be customized for a profession, a vocation, an industry, a technology, an occupation, a business, or other entities having collaborative workspaces. One or more automations in an automation package may also be customized based on specific use cases for certain tasks, such as tracking project progress, enabling communications between remote individuals, or managing files between teams (see at least Mann: ¶ [0603]).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Mann, the results of the combination were predictable.
Maikhuri / Mann system for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Coppersmith in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- partitioning, using the first AI model (see at least Coppersmith: Col. 1, Lns. 63-67 & Col. 2, Lns. 65-67 & Col. 3, Lns. 1-2. Coppersmith notes that Figs. 1a, 1b and 1C, computer system 100 may use NLP and other artificial intelligence techniques to evaluate Slack messages 110 and other free-form text 112, other sensor monitoring 114 of team members and their environment, behavior and communication of workers or team members of a team, and surveys to assess emotional content, sentiment, and psychological well-being of the team members and of the team as a whole.), the text content into the text units (see at least Coppersmith: Col. 3, Lns. 1-7 & Col. 6, Lns. 50-54 & Col. 7, Lns. 46-60. Coppersmith notes that emotion analysis 132 may analyze for emotions such as joy, sadness, surprise, anger and fear and a residual category for messages that express no emotion. See also Coppersmith at Col. 6, Lns. 50-54. Individual data streams such as Slack messages, text messages, or intra-team messages may have captured a large-enough spectrum of team communications to allow system 100 to show mental health, well-being, or stress state.) that are each a minimal body of text containing well-being-related information (see at least Coppersmith: Figs. 6B-6C & Col. 14, Lns. 44-50 & Col. 14, Lns. 60-64 & Col. 15, Lns. 1-11. Coppersmith notes that messages may be grouped by the emotion and/or sentiment. For example, if there are eight categories of emotions (seven emotions plus no emotion) and three sentiments (positive, negative, and neutral), messages may be grouped into twenty-four categories, and messages may be grouped by histogramming message counts into those twenty-four categories. In each time interval, system 100 may collect a count of the total number of messages exchanged in the group. Data may reflect some time interval, such as six months (for example, 60 three-day periods). The scores may be summed to the margins—for example, for a three-day period, the twenty-four cells may be summed into eight emotion margin cells, and three margin sentiment cells. In step 662, as a further data normalization, system 100 may compute a mean over the entire time interval, and then normalize each margin cell to a standard deviation measure. For example, if the 60 cells have an average number of 20 messages per day showing “joy” emotion, with a standard deviation of 3 counts, then cells with 16, 20, and 22 counts would be normalized to “−1.33, 0, 0.67” for further processing.);
- classifying (see at least Coppersmith: Fig. 1 denoting emotional classifiers 132 and sentiment classifiers 134.), using the second AI model (see at least Coppersmith: Col. 1, Lns. 63-67 & Col. 2, Lns. 65-67 & Col. 3, Lns. 1-2. Coppersmith notes that Figs. 1a, 1b and 1C, computer system 100 may use NLP and other artificial intelligence techniques to evaluate Slack messages 110 and other free-form text 112, other sensor monitoring 114 of team members and their environment, behavior and communication of workers or team members of a team, and surveys to assess emotional content, sentiment, and psychological well-being of the team members and of the team as a whole.), the text units into the respective well-being-related categories (see at least Coppersmith: Col. 3, Lns. 1-7 & Col. 6, Lns. 50-54 & Col. 7, Lns. 46-60. Coppersmith notes that emotion analysis 132 may analyze for emotions such as joy, sadness, surprise, anger and fear and a residual category for messages that express no emotion. See also Coppersmith at Col. 6, Lns. 50-54. Individual data streams such as Slack messages, text messages, or intra-team messages may have captured a large-enough spectrum of team communications to allow system 100 to show mental health, well-being, or stress state.)
- generating a time series distribution of the text units among the well-being-related categories based on classifying the text units into the respective well-being-related categories in a continuous and opaque manner that reduces bias (see at least Coppersmith: Figs. 2A – 2G & Col. 5, Lns. 20-44 & Col. 12, Lns. 51-60. Coppersmith teaches that the history of interventions will be recorded alongside the time series data and we will analyze the effect that, e.g., deliberately increasing cross-clique communication and collaboration has on the team's multiplex graph metrics and resulting measured CAL. In this way the system is closed-loop; experimental variations and perturnations may be conducted in as controlled a manner as possible, and effects of those experimental interventions measured to determine viability of future interventions to decrease CAL. See also Coppersmith at Col. 5, Lns. 20-44: “Software 100 may use persistent and passive monitoring of team communications 112, environmental monitoring 114, etc. When monitoring is passive, team members need not act to “turn it on,” which allows for the data collection to be more persistent, reducing sampling error and bias (e.g., participation bias could favor participation from those who have free time to do so, while those team members under higher allostatic load will not as easily make the time).” See also Coppersmith at Figs. 2A-2G noting ““time series distributions” of text units among well-being-related categories.” See also Coppersmith at Col. 3, Lns. 26-36: “The emotional classifiers 132 and sentiment classifiers 134 discerned from text and sensors may include suicide risk, mental illness, joy, fear, sadness, anger, or frustration among team members, team effectiveness, language between two team members (e.g., mentor-mentee, cooperative, consultative, joint problem solving, helpful disagreement, sharp, or condescending), and CAL. These analyses may provide an experimental testbed to evaluate emotional state of individuals and of the team.” See also Coppersmith at Fig. 3C and Fig. 6C as well.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann system for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: partitioning, using the first AI model, the text content into the text units that are each a minimal body of text containing well-being information & classifying, using the second AI model, the text units into the respective well-being-related categories & generating a time series distribution of the text units among the well-being-related categories based on classifying the text units into the respective well-being-related categories in a continuous and opaque manner that reduces bias, and in further view of Coppersmith, whereby artificial intelligence techniques are used in Coppersmith to evaluate Slack messages, other free-form text, other sensor monitoring of team members and their environment, behavior, and communication of workers or team members of a team to assess emotional content, sentiment, and psychological well-being of the team’s members and of the team as a whole (see at least Coppersmith: Col. 2, Lns. 65-67 & Col. 3, Lns. 1-7).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Coppersmith, the results of the combination were predictable.
Maikhuri / Mann / Coppersmith system for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Anders in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- dynamically updating (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. See Anders at ¶ [0025]: “By use of the emotion time series model 160, the emotion service engine 130 dynamically ascertains the state emotion of the user 105 at a certain time”.), by the one or more processors (see at least Anders: Fig. 8 & ¶ [0102] & ¶ [0104-0105]. Anders notes one or more processors 16 shown in the computer system of Fig. 8.) in real-time, the time series distribution based on additional text content received after initial generation of the time series distribution (see at least Anders: ¶ [0039-0040] & Fig. 5 & Fig. 7. Anders notes that the emotion service engine 130 can also modify the calendar of the user 105 by adding additional alerts for an existing schedule or generating a new event on the calendar, with a notice “Don't forget the office outing is this Friday! Two more days to go!” or “The quarterly report is due in two weeks. Please have it ready by next Wednesday for a review before the presentation!”. Anders at ¶ [0040] teaches that the emotion service engine 130 also obtains updates on the emotion time series data 113 from the data collection devices 107. The emotion service engine 130 trains the emotion time series model 160, by machine learning with the updated inputs and/or the feedback 199 from the user 105, in order to improve accuracy of the emotion time series model 160 with the time progression of the baseline as well as the predictions.);
- causing a user interface to be displayed on a user device (see at least Anders: Fig. 8 & ¶ [0110]. Anders notes that in addition to or in place of having external devices 14 and the display 24, which can be configured to provide user interface functionality, computing node 10 can include another display 25 connected to bus 18.), the user interface (see at least Anders: Fig. 8 & ¶ [0110].) including the dynamically updated (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. See Anders at ¶ [0025]: “By use of the emotion time series model 160, the emotion service engine 130 dynamically ascertains the state emotion of the user 105 at a certain time”.) time series distribution of the text units among the well-being-related categories for the organization (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. Anders teaches that the emotion service engine 130 also obtains updates on the emotion time series data 113 from the data collection devices 107. The emotion service engine 130 trains the emotion time series model 160, by machine learning with the updated inputs and/or the feedback 199 from the user 105, in order to improve accuracy of the emotion time series model 160 with the time progression of the baseline as well as the predictions. See also Anders at ¶ [0044]: The emotion service engine 130 processes the collected text contents and classifies the text contents into certain categories indicative of states of emotion, by use of the external natural language understanding tools 170, which includes topic modeling tools. All emotion time series data 113 are respectively associated with a unique time stamp that records the time upon which the emotion time series data 113 had been observed/collected on the user 105. See also Anders at Fig. 5 noting an “Emotion Time Series” 500 and “Basic Emotion Time Series: Joy 510”. See also Anders at Fig. 7 & ¶ [0078]: “The application-emotion time series graph 700 in a first area 701 is for a chat application, which represents a decreasing emotion score over time for the duration between “0:00:00” and “0:18:00”, which indicates that using the chat application program has a negative effect on the emotion score.”).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith system for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: dynamically updating, in real-time, the time series distribution based on additional text content received after initial generation of the time series distribution & causing a user interface to be displayed on a user device, the user interface including the dynamically updated time series distribution of the text units among the well-being-related categories for the organization, and in further view of Anders, whereby the emotion service engine also obtains updates on the emotion time series data. The emotion service engine trains the emotion time series model, by machine learning with the updated inputs and/or the feedback from the user, in order to improve accuracy of the emotion time series model with the time progression of the baseline as well as the predictions. The emotion service engine repeats through when preconfigured conditions for re-processing are met, including, but not limited to, cumulating a predefined amount of input data, obtaining a new instance of the target emotion-time value pair (see at least Anders: ¶ [0040].)
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Anders, the results of the combination were predictable.
Maikhuri / Mann / Coppersmith / Anders system for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Shook in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- retraining (see at least Shook: ¶ [0151] & Fig. 4. Shook teaches that data collected at the data collected 401 can be used to iteratively retrain the model to accurately predict engagement scores.), using the training algorithm (see at least Shook: Figs. 6-10).), the first AI model and/or the second AI model (see at least Shook: Figs. 3-4 & ¶ [0105-0106]. Shook teaches creating a workflow pipeline that uses the combination of above-mentioned data manipulation techniques, that includes identifying inconsistencies in the data, leveraging NLP to correctly parse unstructured data, and using data normalization techniques to standardize the data. Additionally, automated validation checks are be used to ensure high quality of data used for pattern recognition and insights. Natural Language Processing/Understanding (NLP/NLU) techniques are be used to identify understand the semantics on common ground.) using updated training data based on the additional text content (see at least Shook: ¶ [0134-0135] & ¶ [0143] & ¶ [0148] & Figs. 3-4. Shook notes that data Augmentation 445 can be used to generate additional data and diversify the dataset. For example, if there's bias towards positive appraisal feedback, synthetic data points representing negative feedback can be generated to balance the distribution and balance the distribution of occurrence. Data model of the entity data can be defined and re-fined to collect relevant data for the entities that can be used for training the ML model. See also Shook at ¶ [0143]: At 470, the model is tuned for accuracy by implementing calibration and de-biasing techniques that support improved accuracy of the prediction of the engagement scores. Calibration techniques can be used to adjust the model's predictions to align with ground truth outcomes and reduce bias. Such calibration techniques can involve comparing predicted probabilities to observed outcomes and performing adjustments to the model as necessary. See also Shook at Fig. 3 noting “training data 340”.), and adjusting one or more weights of the first AI model and/or the second AI model (see at least Shook: ¶ [0143] & ¶ [0145-0147]. Shook teaches executing the ML model to forecast the set of engagement index score for the set of entities based on the data provided for the set of entities comprises: defining weights for contribution of the parameters to be used for the execution, wherein a weight adjusts a contribution of a parameter when evaluated by the ML model. The weights assigned to different influencing factors as parameters of the ML model can vary based on at least one of organizational priorities, the specific goals, and/or individual employee needs. See also Shook at ¶ [0147]: “Weights assigned to parameters of the ML model can be adjusted to account for such identified bias in the provided data for the ML model execution. By considering a combination of influencing factors on entity level, adjusting weights as needed based on identified bias in provided data for executing the ML model, a centralized source platform that runs the ML model or another system that is configured to execute the ML model can generate tailored-made action plans that are both personalized and equitable for all employees”.) based on a comparison between previous outputs and updated known outcomes (see at least Shook: ¶ [0103] & ¶ [0137-0138] & ¶ [0141-0143]. Shook notes that the obtained training data is data that is enhanced based on the techniques implemented at the data enhancement component 402. The trained model can include defined parameters as previously discussed for which data is collected from the data collector 401. See also Shook at ¶ [0103]: When the training data 340 is generated, considerations for identifying and mitigation hidden bias can be made so that different data augmentation techniques can be performed to modify the collected entity data 320 to remove bias and thus when used for training the ML model 380 to provide a more accurate model that is de-biased compared to other models that can rely on biased data. See also Shook at ¶ [0143]: At 470, the model is tuned for accuracy by implementing calibration and de-biasing techniques that support improved accuracy of the prediction of the engagement scores. Calibration techniques can be used to adjust the model's predictions to align with ground truth outcomes and reduce bias. Such calibration techniques can involve comparing predicted probabilities to observed outcomes and performing adjustments to the model as necessary. See also Shook at ¶ [0148]: Association techniques can be applied that rely on actions that are available to be performed for entities that are determined to be influences by a given influence factor so that their engagement score, when recomputed after the actions are implemented can be modified to show improved engagement compared to the initially computed engagement score.);
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith / Anders system for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: retraining, using the training algorithm, the first AI model and/or the second AI model using updated training data based on the additional text content, and adjusting one or more weights of the first AI model and/or the second AI model based on a comparison between pervious outputs and updated known outcomes, and in further view of Shook, by using advanced algorithms and ML techniques, personalized and curated plans to enhance engagement index scores of employees across an organization can be provided. By proactively addressing specific areas for improvement and aligning them with employee preferences, organizations can foster an engaged and motivated workforce, leading to higher overall organization-wide engagement levels (see at least Shook: ¶ [0102].).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Shook, the results of the combination were predictable.
Regarding Independent Claim 15, Maikhuri non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units teaches the following:
- storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations (see at least Maikhuri: ¶ [0202] & Fig. 14.) comprising:
- training (see at least Maikhuri: Fig. 7 & Fig. 10 & ¶ [0119]. Maikhuri notes training content 136, and the associated analyses with these modules, to help make complex decisions for the user. The PARE 105 is configured to perform personal analysis on the user 113 a (via the user interactions 114 that are recorded and analyzed) and make one or more recommendations (e.g., in the form of first information 130 that includes, but is not limited to, assessments, user reports, recommendations, and/or feedback), with goals of helping a user to understand, analyze, and improve personal and professional user behavior and interactions, to meet personal effectiveness goals.), using a training algorithm (see at least Maikhuri: ¶ [0169] & ¶ [0176-0178]. Maikhuri notes that emotion Classification 812 is the classification algorithm to classify emotions based on the extracted features. The classification has various methods, which classify the images into multiple classes. See also Maikhuri at ¶ [0111]: “This intelligent and specifically trained IPAE 102 (and optional associated robotic automation is) configured to understand the employees' context, role, domain, intent, and communication/task.” See also Maikhuri at ¶ [0176-0178]: “This model uses historical training data containing multi-dimension data points to train the model. Once the model is fully trained, the conversation's state (intent, sentiment, context) is passed to predict the following best action. The algorithm Random Forest uses a large group of complex decision trees and can provide classification predictions with a high degree of accuracy on any size of data. This engine algorithm will predict the recommended virtual assistant with the accuracy or likelihood percentage.”.), a first artificial intelligence (AI) model to partition text content into text units (see at least Maikhuri: Fig. 1A & ¶ [0116] & ¶ [0128] & ¶ [0162]. Maikhuri teaches that the text/natural language processing module 104 is configured for entity recognition and content classification, as well as intent and sentiment analysis. The text/natural language processing module 104 also is configured to cooperate with outputs of the voice/audio analysis module 106 and video analysis module 108 to derive and/or determine intent and sentiment from that content, as well as to assist in performing content classification. See also Maikhuri at ¶ [0027]: “The first analysis configured to analyze the set of raw data records for at least one of sentiments, emotions, and intent; performing a second analysis on the set of raw data records, the second analysis configured to segment the set of raw data records.” See also Maikhuri at ¶ [0128]: The second analysis may use automatic text segmentation 166, which is configured to breaks up text into topically-consistent segments. Text segmentation, as is known, is related to natural language processing, document classification, and information retrieval. In block 165, using text segmentation, specific features can be extracted and segmented in the data, such as determinations that, within the raw user data 144, a specific project, context, individual, etc., is mentioned, or that a specific string of words are related and together represent an actionable action or statement. See also Maikhuri at ¶ [0162]: “The artificial intelligence (AI) 422 further processes this information.”) and a second AI model (see at least Maikhuri: ¶ [0092] & ¶ [0115] & ¶ [0126]. Maikhuri teaches that the analysis of the intelligent processing engine 101 uses Natural Language Processing (NLP) to understand the communication context, intent, sentiment, etc., and sends these details to the personal knowledge/domain expertise repository. A part of this functionality is the IPAE's ability to understand and express itself in natural language, creating an intuitive way for the employee/person to perform tasks.) to classify the text units into respective well-being-related categories (see at least Maikhuri: ¶ [0126] & ¶ [0164]. Maikhuri teaches that sentiment analysis a process of using automated processes and/or machine learning, such as natural language processing (NLP), text analysis, and statistics to analyze the sentiment in one or more of string of words (such as an email, a social media post, a statement, a text, etc.). Sentiment analysis includes techniques and methods for understanding emotions by use of software. Natural language processing, statistics, and text analysis are used as part of sentiment analysis to extract, and identify the sentiment of words (e.g., to determine if words may be positive, negative, neutral, and/or whether there are additional emotions that can be inferred, such as anger, confusion, enthusiasm, humor, etc.). See also Maikhuri at ¶ [0147]: “A general-purpose entity can correspond to a storage location having an identifier that may be common to many users, such as “to-do list,” “summary of meetings,” “feedback from boss,” “overdue tasks.” When raw user data is analyzed, the intelligent processing engine 101 may determine that, based on the processed and interpreted content, a given raw user data record may fit into one of the predetermined entity categories.” See also Maikhuri at ¶ [0164]: Sentiment analysis is done through text data. Audio data also is processed to help detect a person's emotions just by their voice which will help to know and interpret that person's actions and/or their behavior. Neural network techniques such as multilayer perceptron (MLP) and long short-term memory (LSTM) are less advantageous, so techniques such as convolutional neural network (CNN) are used to classify in the problem/situation where different emotions need to be categorized.) based on training data including inputs and known outcomes (see at least Maikhuri: ¶ [0157] & ¶ [0177] & Fig. 7 & Fig. 10. Maikhuri notes that the IPAE 102, after analyzing the communication, can dynamically create first information 130 such as an assessment summary that includes specific recommendations, tasks, or brief summaries 202 of all messages and chats that day, so that the user 113 a does not have to go back and review multiple threads or chats for final outcome and actions. See also Maikhuri at ¶ [0112-0114]: User interactions inputs, and related user interaction data, including inputs input as speech/text and audio and sends/converts output as speech/text/audio, as applicable. For example, the communications gateway 103 can receive information in text, video, and/or audio format. See also Maikhuri at ¶ [0129]: The IPAE 102 classifies received raw user data 144 (e.g., set of user data, user interactions, inputs, and optional related user interaction data) to organize by sentiment, emotion, intent, domain, project, environment, task, interaction group, etc.)
Maikhuri non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Mann in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- receiving, from one or more applications of a digital workspace of an organization (see at least Mann: ¶ [0032] & ¶ [0434] & ¶ [0603]. Mann notes tracking electronic connections between a plurality of entities in an electronic workspace; tracking characteristics of the electronic connections between the plurality of entities in the electronic workspace. See also Mann at ¶ [0434]: Workspace templates icon 2614 may enable users to package boards and dashboards as a unified solution. See also Mann at ¶ [0603]: One or more automations in an automation package may be customized for a profession, a vocation, an industry, a technology, an occupation, a business, or other entities having collaborative workspaces. One or more automations in an automation package may also be customized based on specific use cases for certain tasks, such as tracking project progress, enabling communications between remote individuals, or managing files between teams. See also Mann at ¶ [0636]: A collaborator may use a primary application as a main working environment, while functionality of third-party applications may be integrated into the primary application. See also Mann at Fig. 141 of the digital workspaces.), the text content generated by users of the organization (see at least Mann: ¶ [0549] & ¶ [1396]. Mann notes characterizing text messages may include recording key words from the text messages. The system may analyze the logged chat messages from a communication and determine key words or phrases spoken by each individual or all individuals during the video or audio communication. This may enable users to track action items at the conclusion of the communication. The system may enable participants to manually mark key words from the communication or text messages for recordation so that the participants may later refer to the key word. See also Mann at ¶ [1396]: Mann notes that semantic analysis involves a computer process for drawing meaning from text. It may involve identifying relationships between individual words in a particular context within sentences, paragraphs, or whole documents by electronically analyzing grammatical structure and identifying relationships between particular words in a particular context. After semantic analysis is performed on first data in a first board and second data in a second board, the at least one processor can compare the results to determine a similarity.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: receiving, from one or more applications of a digital workspace of an organization, the text content generated by users of the organization, and in view of Mann, whereby the relationship or connection may be established by linking the automation and the table, or by assigning a common code, address, or other designation to the automation and the table. One or more automations in an automation package may be customized for a profession, a vocation, an industry, a technology, an occupation, a business, or other entities having collaborative workspaces. One or more automations in an automation package may also be customized based on specific use cases for certain tasks, such as tracking project progress, enabling communications between remote individuals, or managing files between teams (see at least Mann: ¶ [0603]).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Mann, the results of the combination were predictable.
Maikhuri / Mann non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Coppersmith in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- partitioning, using the first AI model (see at least Coppersmith: Col. 1, Lns. 63-67 & Col. 2, Lns. 65-67 & Col. 3, Lns. 1-2. Coppersmith notes that Figs. 1a, 1b and 1C, computer system 100 may use NLP and other artificial intelligence techniques to evaluate Slack messages 110 and other free-form text 112, other sensor monitoring 114 of team members and their environment, behavior and communication of workers or team members of a team, and surveys to assess emotional content, sentiment, and psychological well-being of the team members and of the team as a whole.), the text content into the text units (see at least Coppersmith: Col. 3, Lns. 1-7 & Col. 6, Lns. 50-54 & Col. 7, Lns. 46-60. Coppersmith notes that emotion analysis 132 may analyze for emotions such as joy, sadness, surprise, anger and fear and a residual category for messages that express no emotion. See also Coppersmith at Col. 6, Lns. 50-54. Individual data streams such as Slack messages, text messages, or intra-team messages may have captured a large-enough spectrum of team communications to allow system 100 to show mental health, well-being, or stress state.) that are each a minimal body of text containing well-being-related information (see at least Coppersmith: Figs. 6B-6C & Col. 14, Lns. 44-50 & Col. 14, Lns. 60-64 & Col. 15, Lns. 1-11. Coppersmith notes that messages may be grouped by the emotion and/or sentiment. For example, if there are eight categories of emotions (seven emotions plus no emotion) and three sentiments (positive, negative, and neutral), messages may be grouped into twenty-four categories, and messages may be grouped by histogramming message counts into those twenty-four categories. In each time interval, system 100 may collect a count of the total number of messages exchanged in the group. Data may reflect some time interval, such as six months (for example, 60 three-day periods). The scores may be summed to the margins—for example, for a three-day period, the twenty-four cells may be summed into eight emotion margin cells, and three margin sentiment cells. In step 662, as a further data normalization, system 100 may compute a mean over the entire time interval, and then normalize each margin cell to a standard deviation measure. For example, if the 60 cells have an average number of 20 messages per day showing “joy” emotion, with a standard deviation of 3 counts, then cells with 16, 20, and 22 counts would be normalized to “−1.33, 0, 0.67” for further processing.);
- classifying (see at least Coppersmith: Fig. 1 denoting emotional classifiers 132 and sentiment classifiers 134.), using the second AI model (see at least Coppersmith: Col. 1, Lns. 63-67 & Col. 2, Lns. 65-67 & Col. 3, Lns. 1-2. Coppersmith notes that Figs. 1a, 1b and 1C, computer system 100 may use NLP and other artificial intelligence techniques to evaluate Slack messages 110 and other free-form text 112, other sensor monitoring 114 of team members and their environment, behavior and communication of workers or team members of a team, and surveys to assess emotional content, sentiment, and psychological well-being of the team members and of the team as a whole.), the text units into the respective well-being-related categories (see at least Coppersmith: Col. 3, Lns. 1-7 & Col. 6, Lns. 50-54 & Col. 7, Lns. 46-60. Coppersmith notes that emotion analysis 132 may analyze for emotions such as joy, sadness, surprise, anger and fear and a residual category for messages that express no emotion. See also Coppersmith at Col. 6, Lns. 50-54. Individual data streams such as Slack messages, text messages, or intra-team messages may have captured a large-enough spectrum of team communications to allow system 100 to show mental health, well-being, or stress state.)
- generating a time series distribution of the text units among the well-being-related categories based on classifying the text units into the respective well-being-related categories in a continuous and opaque manner that reduces bias (see at least Coppersmith: Figs. 2A – 2G & Col. 5, Lns. 20-44 & Col. 12, Lns. 51-60. Coppersmith teaches that the history of interventions will be recorded alongside the time series data and we will analyze the effect that, e.g., deliberately increasing cross-clique communication and collaboration has on the team's multiplex graph metrics and resulting measured CAL. In this way the system is closed-loop; experimental variations and perturnations may be conducted in as controlled a manner as possible, and effects of those experimental interventions measured to determine viability of future interventions to decrease CAL. See also Coppersmith at Col. 5, Lns. 20-44: “Software 100 may use persistent and passive monitoring of team communications 112, environmental monitoring 114, etc. When monitoring is passive, team members need not act to “turn it on,” which allows for the data collection to be more persistent, reducing sampling error and bias (e.g., participation bias could favor participation from those who have free time to do so, while those team members under higher allostatic load will not as easily make the time).” See also Coppersmith at Figs. 2A-2G noting ““time series distributions” of text units among well-being-related categories.” See also Coppersmith at Col. 3, Lns. 26-36: “The emotional classifiers 132 and sentiment classifiers 134 discerned from text and sensors may include suicide risk, mental illness, joy, fear, sadness, anger, or frustration among team members, team effectiveness, language between two team members (e.g., mentor-mentee, cooperative, consultative, joint problem solving, helpful disagreement, sharp, or condescending), and CAL. These analyses may provide an experimental testbed to evaluate emotional state of individuals and of the team.” See also Coppersmith at Fig. 3C and Fig. 6C as well.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: partitioning, using the first AI model, the text content into the text units that are each a minimal body of text containing well-being information & classifying, using the second AI model, the text units into the respective well-being-related categories & generating a time series distribution of the text units among the well-being-related categories based on classifying the text units into the respective well-being-related categories in a continuous and opaque manner that reduces bias, and in further view of Coppersmith, whereby artificial intelligence techniques are used in Coppersmith to evaluate Slack messages, other free-form text, other sensor monitoring of team members and their environment, behavior, and communication of workers or team members of a team to assess emotional content, sentiment, and psychological well-being of the team’s members and of the team as a whole (see at least Coppersmith: Col. 2, Lns. 65-67 & Col. 3, Lns. 1-7).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Coppersmith, the results of the combination were predictable.
Maikhuri / Mann / Coppersmith non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Anders in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- dynamically updating (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. See Anders at ¶ [0025]: “By use of the emotion time series model 160, the emotion service engine 130 dynamically ascertains the state emotion of the user 105 at a certain time”.), by the one or more processors (see at least Anders: Fig. 8 & ¶ [0102] & ¶ [0104-0105]. Anders notes one or more processors 16 shown in the computer system of Fig. 8.) in real-time, the time series distribution based on additional text content received after initial generation of the time series distribution (see at least Anders: ¶ [0039-0040] & Fig. 5 & Fig. 7. Anders notes that the emotion service engine 130 can also modify the calendar of the user 105 by adding additional alerts for an existing schedule or generating a new event on the calendar, with a notice “Don't forget the office outing is this Friday! Two more days to go!” or “The quarterly report is due in two weeks. Please have it ready by next Wednesday for a review before the presentation!”. Anders at ¶ [0040] teaches that the emotion service engine 130 also obtains updates on the emotion time series data 113 from the data collection devices 107. The emotion service engine 130 trains the emotion time series model 160, by machine learning with the updated inputs and/or the feedback 199 from the user 105, in order to improve accuracy of the emotion time series model 160 with the time progression of the baseline as well as the predictions.);
- causing a user interface to be displayed on a user device (see at least Anders: Fig. 8 & ¶ [0110]. Anders notes that in addition to or in place of having external devices 14 and the display 24, which can be configured to provide user interface functionality, computing node 10 can include another display 25 connected to bus 18.), the user interface (see at least Anders: Fig. 8 & ¶ [0110].) including the dynamically updated (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. See Anders at ¶ [0025]: “By use of the emotion time series model 160, the emotion service engine 130 dynamically ascertains the state emotion of the user 105 at a certain time”.) time series distribution of the text units among the well-being-related categories for the organization (see at least Anders: ¶ [0040] & ¶ [0044] & Fig. 5. Anders teaches that the emotion service engine 130 also obtains updates on the emotion time series data 113 from the data collection devices 107. The emotion service engine 130 trains the emotion time series model 160, by machine learning with the updated inputs and/or the feedback 199 from the user 105, in order to improve accuracy of the emotion time series model 160 with the time progression of the baseline as well as the predictions. See also Anders at ¶ [0044]: The emotion service engine 130 processes the collected text contents and classifies the text contents into certain categories indicative of states of emotion, by use of the external natural language understanding tools 170, which includes topic modeling tools. All emotion time series data 113 are respectively associated with a unique time stamp that records the time upon which the emotion time series data 113 had been observed/collected on the user 105. See also Anders at Fig. 5 noting an “Emotion Time Series” 500 and “Basic Emotion Time Series: Joy 510”. See also Anders at Fig. 7 & ¶ [0078]: “The application-emotion time series graph 700 in a first area 701 is for a chat application, which represents a decreasing emotion score over time for the duration between “0:00:00” and “0:18:00”, which indicates that using the chat application program has a negative effect on the emotion score.”).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: dynamically updating, in real-time, the time series distribution based on additional text content received after initial generation of the time series distribution & causing a user interface to be displayed on a user device, the user interface including the dynamically updated time series distribution of the text units among the well-being-related categories for the organization, and in further view of Anders, whereby the emotion service engine also obtains updates on the emotion time series data. The emotion service engine trains the emotion time series model, by machine learning with the updated inputs and/or the feedback from the user, in order to improve accuracy of the emotion time series model with the time progression of the baseline as well as the predictions. The emotion service engine repeats through when preconfigured conditions for re-processing are met, including, but not limited to, cumulating a predefined amount of input data, obtaining a new instance of the target emotion-time value pair (see at least Anders: ¶ [0040].)
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Anders, the results of the combination were predictable.
Maikhuri / Mann / Coppersmith / Anders non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units does not explicitly disclose, but Shook in the analogous art for using artificial intelligence models to generate a time series distribution of text units teaches the following limitations:
- retraining (see at least Shook: ¶ [0151] & Fig. 4. Shook teaches that data collected at the data collected 401 can be used to iteratively retrain the model to accurately predict engagement scores.), using the training algorithm (see at least Shook: Figs. 6-10).), the first AI model and/or the second AI model (see at least Shook: Figs. 3-4 & ¶ [0105-0106]. Shook teaches creating a workflow pipeline that uses the combination of above-mentioned data manipulation techniques, that includes identifying inconsistencies in the data, leveraging NLP to correctly parse unstructured data, and using data normalization techniques to standardize the data. Additionally, automated validation checks are be used to ensure high quality of data used for pattern recognition and insights. Natural Language Processing/Understanding (NLP/NLU) techniques are be used to identify understand the semantics on common ground.) using updated training data based on the additional text content (see at least Shook: ¶ [0134-0135] & ¶ [0143] & ¶ [0148] & Figs. 3-4. Shook notes that data Augmentation 445 can be used to generate additional data and diversify the dataset. For example, if there's bias towards positive appraisal feedback, synthetic data points representing negative feedback can be generated to balance the distribution and balance the distribution of occurrence. Data model of the entity data can be defined and re-fined to collect relevant data for the entities that can be used for training the ML model. See also Shook at ¶ [0143]: At 470, the model is tuned for accuracy by implementing calibration and de-biasing techniques that support improved accuracy of the prediction of the engagement scores. Calibration techniques can be used to adjust the model's predictions to align with ground truth outcomes and reduce bias. Such calibration techniques can involve comparing predicted probabilities to observed outcomes and performing adjustments to the model as necessary. See also Shook at Fig. 3 noting “training data 340”.), and adjusting one or more weights of the first AI model and/or the second AI model (see at least Shook: ¶ [0143] & ¶ [0145-0147]. Shook teaches executing the ML model to forecast the set of engagement index score for the set of entities based on the data provided for the set of entities comprises: defining weights for contribution of the parameters to be used for the execution, wherein a weight adjusts a contribution of a parameter when evaluated by the ML model. The weights assigned to different influencing factors as parameters of the ML model can vary based on at least one of organizational priorities, the specific goals, and/or individual employee needs. See also Shook at ¶ [0147]: “Weights assigned to parameters of the ML model can be adjusted to account for such identified bias in the provided data for the ML model execution. By considering a combination of influencing factors on entity level, adjusting weights as needed based on identified bias in provided data for executing the ML model, a centralized source platform that runs the ML model or another system that is configured to execute the ML model can generate tailored-made action plans that are both personalized and equitable for all employees”.) based on a comparison between previous outputs and updated known outcomes (see at least Shook: ¶ [0103] & ¶ [0137-0138] & ¶ [0141-0143]. Shook notes that the obtained training data is data that is enhanced based on the techniques implemented at the data enhancement component 402. The trained model can include defined parameters as previously discussed for which data is collected from the data collector 401. See also Shook at ¶ [0103]: When the training data 340 is generated, considerations for identifying and mitigation hidden bias can be made so that different data augmentation techniques can be performed to modify the collected entity data 320 to remove bias and thus when used for training the ML model 380 to provide a more accurate model that is de-biased compared to other models that can rely on biased data. See also Shook at ¶ [0143]: At 470, the model is tuned for accuracy by implementing calibration and de-biasing techniques that support improved accuracy of the prediction of the engagement scores. Calibration techniques can be used to adjust the model's predictions to align with ground truth outcomes and reduce bias. Such calibration techniques can involve comparing predicted probabilities to observed outcomes and performing adjustments to the model as necessary. See also Shook at ¶ [0148]: Association techniques can be applied that rely on actions that are available to be performed for entities that are determined to be influences by a given influence factor so that their engagement score, when recomputed after the actions are implemented can be modified to show improved engagement compared to the initially computed engagement score.);
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith / Anders non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: retraining, using the training algorithm, the first AI model and/or the second AI model using updated training data based on the additional text content, and adjusting one or more weights of the first AI model and/or the second AI model based on a comparison between pervious outputs and updated known outcomes, and in further view of Shook, by using advanced algorithms and ML techniques, personalized and curated plans to enhance engagement index scores of employees across an organization can be provided. By proactively addressing specific areas for improvement and aligning them with employee preferences, organizations can foster an engaged and motivated workforce, leading to higher overall organization-wide engagement levels (see at least Shook: ¶ [0102].).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Shook, the results of the combination were predictable.
Regarding Dependent Claims 2, 9 and 16, Maikhuri / Mann / Coppersmith / Anders / Shook method / system / non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units teaches the limitations of Independent Claims 1, 8 and 15 above, and Mann further teaches the method / system / non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units comprising:
- wherein receiving the text content comprises receiving the text content from one or more application programming interfaces (APIs) of the one or more applications (see at least Mann: ¶ [0539] & ¶ [0665] & ¶ [1235] & ¶ [1348]. Mann teaches linking of the primary application to the third-party applications may be accomplished using, for example, an Application Programing Interface (API). See also Mann at ¶ [0539]. A conversation duration may include the length of time of the communication or the length of times each participant participated in the communication. A list of key words spoken in the communication may obtained via speech recognition software, such as a speech to text API, or any other suitable mechanism for deriving text from speech. See also Mann at ¶ [1235]. Mann notes that a mediator overlay may include a webhook, web callback, HTTP push API, or any other method that may provide information or data from other applications in real-time. See also Mann at ¶ [1348]. Mann notes that it may contain one or more functions (e.g., Application Programming Interfaces (APIs)) for accessing one or more audio files from an application or other electronic system.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith / Anders / Shook method / system / non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: wherein receiving the text content comprises receiving the text content from one or more application programming interfaces (APIs) of the one or more applications, and in further view of Mann, whereby the data structure may, for example, include one or more audio files and corresponding identifications for looking up the one or more audio files; or it may contain one or more functions (e.g., Application Programming Interfaces (APIs)) for accessing one or more audio files from an application or other electronic system (see at least Mann: ¶ [1348].). Also a user may desire, for example, to utilize automatic processes to generate graphical representations based on user selections of table data, without having to manually create graphical representations and without having to interact with a separate graphical user interface. In addition, the disclosed computerized systems and methods may generate a link between the table data and the graphical representation, thereby leading to real-time or near real-time updates of the graphical representation as a result of changes in the table data. This provides several benefits over extant processes that rely on manual updating or other user-dependent input to update graphical representations, resulting in saved time (see at least Mann: ¶ [0770].).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Mann, the results of the combination were predictable.
Regarding Dependent Claims 3, 10 and 17, Maikhuri / Mann / Coppersmith / Anders / Shook method / system / non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units teaches the limitations of Independent Claims 1, 8 and 15 above, and Maikhuri further teaches the method / system / non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units comprising:
- wherein the text units are sentences (see at least Maikhuri: ¶ [0179] & ¶ [0197]. Maikhuri teaches that corpus data contains the words and phrases and the intent associated with each sentence. Also, that tokenization 1108 of the sentences is done by extracting tokens(terms/words) from the corpus.)
Regarding Dependent Claims 4, 11 and 18, Maikhuri / Mann / Coppersmith / Anders / Shook method / system / non-transitory computer readable medium for using artificial intelligence models to generate a time series distribution of text units teaches the limitations of Independent Claims 1, 8 and 15 above, and Maikhuri further teaches the method / system / non-transitory computer readable medium for using artificial intelligence models to generate a time series distribution of text units comprising:
- storing, by the one or more processors (see at least Maikhuri: ¶ [0204-0205].), information identifying the text units, respective users associated with the text units, and respective categories to which the text units are classified in a database (see at least Maikhuri: ¶ [0121-0123] & ¶ [0155] & ¶ [0164]. Maikhuri notes “storing/retrieval of various expression, actions, applications in the daily conversations, decisions, and task of a user”. See also Maikhuri at ¶ [0121]: Storing/retrieval of various expressions, actions, applications in the daily conversations, decisions, task of a user and build the contexts and semantics based on the channel (e.g., type, such as email, messaging, phone calls/audio, video calls and meetings, documents produced, etc.) of the conversation or interaction that takes place in relations with a user, content and its associated sentiments, for efficient processing (storage and retrieval) of knowledge about and for that user. See also Maikhuri at ¶ [0123]: The repository 110 can be configured to store each user's facial expression in their respective segment data. The text and voice are analyzed and stored in text with context, domain, time, and person classification. See also Maikhuri at ¶ [0147]: A given raw user data record may fit into one of the predetermined entity categories. A domain-specific entity can correspond to a storage location having an identifier that is specific to a user's role, employer, co-worker name, location, assigned project, etc., such as “Project Mars-Silo,” or “emails with James.” This accumulated information enables the IPAE 102 to better work with the user, emulate the user, and/or take actions on behalf of the user. See also Maikhuri at ¶ [0155]: The repository 110 can classify stored information based on various features 204, including but not limited to date and time, channel (e.g., did conversation take place via chat, at a meeting, over email, etc.), type of conversation, text and details of conversation, format of the information (text, audio, video, etc.)). See also Maikhuri at ¶ [0164]: Techniques such as convolutional neural network (CNN) are used to classify in the problem/situation where different emotions need to be categorized.)
Regarding Dependent Claims 5, 12 and 19, Maikhuri / Mann / Coppersmith / Anders / Shook method / system / non-transitory computer readable medium for using artificial intelligence models to generate a time series distribution of text units teaches the limitations of Claims 1, 4, 8, 11, 15 and 18 above, and Coppersmith further teaches the method / system / non-transitory computer readable medium for using artificial intelligence models to generate a time series distribution of text units comprising:
- receiving, by the one or more processors (see at least Coppersmith: Fig. 1a & Col. 16, Lns. 54-60. Coppersmith noting a processor shown in Fig. 1a.), a user input identifying a set of user identifiers (see at least Coppersmith: Fig. 4 & Col. 11, Lns. 42-50 & Col. 12, Lns. 22-25.), a time frame (see at least Coppersmith: Fig. 4 & Col. 14, Lns. 54-64.), and a set of applications (see at least Coppersmith: Fig. 1a & Fig. 4.);
- retrieving, by the one or more processors (see at least Coppersmith: Fig. 1a & Col. 16, Lns. 54-60. Coppersmith noting a processor shown in Fig. 1a.) and from the database (see at least Coppersmith: Col. 17, Lns. 21-27.), the information identifying the text units, the respective users associated with the text units, and the respective categories to which the text units are classified (see at least Coppersmith: Fig. 4 & Fig. 6c.), based on receiving the user input identifying the set of user identifiers (see at least Coppersmith: Fig. 4 & Col. 11, Lns. 42-50 & Col. 12, Lns. 22-25.), the time frame (see at least Coppersmith: Fig. 4 & Col. 14, Lns. 54-64.), and the set of applications (see at least Coppersmith: Fig. 1a & Fig. 4.);
- wherein the time series distribution of the text units among the well-being-related categories (see at least Coppersmith: Figs. 2a- 2g.) is generated based on retrieving, from the database (see at least Coppersmith: Col. 17, Lns. 21-27.), the information identifying the text units, the respective users associated with the text units, and the respective categories to which the text units are classified (see at least Coppersmith: Fig. 4 & Fig. 6c.).
It would have been obvious for one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the teachings of Maikhuri / Mann / Coppersmith / Anders / Shook method / system / non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units with the aforementioned teachings of: receiving, by the one or more processors, a user input identifying a set of user identifiers, a time frame, and a set of applications; retrieving, by the one or more processors and from the database, the information identifying the text units, the respective users associated with the text units, and the respective categories to which the text units are classified, based on receiving the user input identifying the set of user identifiers, the time frame, and the set of applications, wherein the time series distribution of the text units among the well-being-related categories is generated based on retrieving, from the database, the information identifying the text units, the respective users associated with the text units, and the respective categories to which the text units are classified, and in further view of Coppersmith, whereby artificial intelligence techniques are used in Coppersmith to evaluate Slack messages, other free-form text, other sensor monitoring of team members and their environment, behavior, and communication of workers or team members of a team to assess emotional content, sentiment, and psychological well-being of the team’s members and of the team as a whole (see at least Coppersmith: Col. 2, Lns. 65-67 & Col. 3, Lns. 1-7).
Further, the claimed invention is merely a combination of old elements in a similar field for using artificial intelligence models to generate a time series distribution of text units, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that, given the existing technical ability to combine the elements as evidenced by Coppersmith, the results of the combination were predictable.
Regarding Dependent Claims 6 and 13, Maikhuri / Mann / Coppersmith / Anders / Shook method / system for using artificial intelligence models to generate a time series distribution of text units teaches the limitations of Independent Claims 1 and 8 above, and Maikhuri further teaches the method / system for using artificial intelligence models to generate a time series distribution of text units comprising:
- wherein the first AI model is an entity recognition model (see at least Maikhuri: ¶ [0116] & ¶ [0163] & Fig. 4. Maikhuri notes that an important and central component of the IPAE 102 is the text/natural language processing module 104, which provides named entity recognition, intent, and sentiment analysis. The text/natural language processing module 104 is configured for entity recognition and content classification, as well as intent and sentiment analysis. See also artificial intelligence (AI) 422 shown in Fig. 4 and “artificial intelligence (AI) powered personal effectiveness management shown in Fig. 1A”.) that is trained to partition the text content into the text units (see at least Maikhuri: ¶ [0128]. Maikhuri notes that the second analysis may use automatic text segmentation 166, which is configured to breaks up text into topically-consistent segments. Text segmentation, as is known, is related to natural language processing, document classification, and information retrieval. In block 165, using text segmentation, specific features can be extracted and segmented in the data, such as determinations that, within the raw user data 144, a specific project, context, individual, etc., is mentioned, or that a specific string of words are related and together represent an actionable action or statement.)
Regarding Dependent Claims 7 and 14, Maikhuri / Mann / Coppersmith / Anders / Shook method / system for using artificial intelligence models to generate a time series distribution of text units teaches the limitations of Independent Claims 1 and 8 above, and Maikhuri further teaches the method / system for using artificial intelligence models to generate a time series distribution of text units comprising:
- wherein the second AI model is a natural language understanding model that is trained to classify the text units into the respective well-being-related categories (see at least Maikhuri: ¶ [0114] & ¶ [0126-0127] & ¶ [0164]. Maikhuri notes that sentiment analysis a process of using automated processes and/or machine learning, such as natural language processing (NLP), text analysis, and statistics to analyze the sentiment in one or more of string of words (such as an email, a social media post, a statement, a text, etc.). Sentiment analysis includes techniques and methods for understanding emotions by use of software. Natural language processing, statistics, and text analysis are used as part of sentiment analysis to extract, and identify the sentiment of words (e.g., to determine if words may be positive, negative, neutral, and/or whether there are additional emotions that can be inferred, such as anger, confusion, enthusiasm, humor, etc.). Supervisors in the non-virtual world have regular opportunities to formally and informally view and assess their employees personal effectiveness and well-being. See also Maikhuri at ¶ [0114]: Natural language processing module 104) and the personal knowledge/domain expertise repository 110, to help ensure that the raw user data 144 is securely pre-processed, analyzed and classified. See also Maikhuri at ¶ [0127]: Text Sentiment Analysis, as noted above, is a capability that uses natural language understanding (NLU) and neural networks to analyze the message and classify the Intent. Sentiment analysis is important in understanding the message context and making appropriate decisions in a priority manner. See also Maikhuri at ¶ [0164]: Techniques such as convolutional neural network (CNN) are used to classify in the problem/situation where different emotions need to be categorized.).
Regarding Dependent Claim 20, Maikhuri / Mann / Coppersmith / Anders / Shook non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units teaches the limitations of Independent Claim 15 above, and Maikhuri further teaches the non-transitory computer-readable medium for using artificial intelligence models to generate a time series distribution of text units comprising:
- wherein the first AI model is an entity recognition model (see at least Maikhuri: ¶ [0116] & ¶ [0163] & Fig. 4. Maikhuri notes that an important and central component of the IPAE 102 is the text/natural language processing module 104, which provides named entity recognition, intent, and sentiment analysis. The text/natural language processing module 104 is configured for entity recognition and content classification, as well as intent and sentiment analysis. See also artificial intelligence (AI) 422 shown in Fig. 4 and “artificial intelligence (AI) powered personal effectiveness management shown in Fig. 1A”.) that is trained to partition the text content into the text units (see at least Maikhuri: ¶ [0128]. Maikhuri notes that the second analysis may use automatic text segmentation 166, which is configured to breaks up text into topically-consistent segments. Text segmentation, as is known, is related to natural language processing, document classification, and information retrieval. In block 165, using text segmentation, specific features can be extracted and segmented in the data, such as determinations that, within the raw user data 144, a specific project, context, individual, etc., is mentioned, or that a specific string of words are related and together represent an actionable action or statement.), and wherein the second AI model is a natural language understanding model that is trained to classify the text units into the respective well-being-related categories (see at least Maikhuri: ¶ [0114] & ¶ [0126-0127] & ¶ [0164]. Maikhuri notes that sentiment analysis a process of using automated processes and/or machine learning, such as natural language processing (NLP), text analysis, and statistics to analyze the sentiment in one or more of string of words (such as an email, a social media post, a statement, a text, etc.). Sentiment analysis includes techniques and methods for understanding emotions by use of software. Natural language processing, statistics, and text analysis are used as part of sentiment analysis to extract, and identify the sentiment of words (e.g., to determine if words may be positive, negative, neutral, and/or whether there are additional emotions that can be inferred, such as anger, confusion, enthusiasm, humor, etc.). Supervisors in the non-virtual world have regular opportunities to formally and informally view and assess their employees personal effectiveness and well-being. See also Maikhuri at ¶ [0114]: Natural language processing module 104) and the personal knowledge/domain expertise repository 110, to help ensure that the raw user data 144 is securely pre-processed, analyzed and classified. See also Maikhuri at ¶ [0127]: Text Sentiment Analysis, as noted above, is a capability that uses natural language understanding (NLU) and neural networks to analyze the message and classify the Intent. Sentiment analysis is important in understanding the message context and making appropriate decisions in a priority manner. See also Maikhuri at ¶ [0164]: Techniques such as convolutional neural network (CNN) are used to classify in the problem/situation where different emotions need to be categorized.).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DERICK HOLZMACHER whose telephone number is (571) 270-7853. The examiner can normally be reached on Monday-Friday 9:00 AM – 6:30 PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, Applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Brian Epstein can be reached on 571-270-5389. The fax phone number for the organization where this application or proceeding is assigned is 571-270-8853.
Information regarding the status of an application may be obtained from Patent Center. Status information for published applications may be obtained from Patent Center. Status information for unpublished applications is available through Patent Center for authorized users only. Should you have questions about access to Patent Center, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/DERICK J HOLZMACHER/ Patent Examiner, Art Unit 3625A
/BRIAN M EPSTEIN/Supervisory Patent Examiner, Art Unit 3625