DETAILED ACTION
Status of Claims
The Response filed 11/25/2025 has been acknowledged. Claims 1, 10, 19 are amended. Claims 2, 4, 11, 13 are cancelled. Claims 1, 3, 5-10, 13, 14-19 are currently pending and have been examined.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 103
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1, 3, 5-8, 10, 12, 14-17, 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Podgorny et al. (US 20210133581 A1) (hereafter Podgorny), in view of Cheng et al. (US 20180121929 A1) (hereafter Cheng), in view of Douglas (US 20200065825 A1) (hereafter Douglas), in view of Jungmeisteris et al. (US 20220374956 A1) (hereafter Jungmeisteris).
As per claim 1:
A computer system for dynamically providing predictive context aware solutions on computing devices to online customers, the computer system comprising:
a processor configured to execute instructions;
a non-transient computer-readable medium comprising instructions that when executed by the processor cause the processor to:
track customer attributes comprising: an online customer behaviour of a particular customer of an entity when interacting with a computer application associated with the entity and on a computer device, the online customer behaviour tracked comprising: a particular flow of navigational events when browsing the computer application indicative of the flow (See Podgorny ¶0036, “Turning to FIG. 2, a problem decoder (200), in accordance with one or more embodiments, is shown. The problem decoder (200) includes a user input encoder (210), a clickstream encoder (220), a shared latent space representation (230), and a shared latent space decoder (240). In combination, these components may output a problem summary (242) from the modalities (202), provided as inputs. The modalities may include the interaction (204) between the user and the support agent, the clickstream (206), and optionally user attributes (208). Other additional components, for example, as shown in FIG. 7, may exist and may be used to update components of the problem decoder (200) using a reinforcement learning paradigm.” Podgorny discloses tracking consumer behavior including navigation information through an application.)
including a sequence of interactions online, order and type of previous pages visited and failed attempts while navigating online in a browsing sequence; (See Podgorny ¶0029, “In one or more embodiments, the application back-end (140) receives a user input provided by the user via the input interface (122) and generates a clickstream (146) from the user input. Broadly speaking, the clickstream (146) may be generated by the application back-end (140) as the user is navigating through the software application. The clickstream (146) may document any type of interaction of the user with the software application. For example, the clickstream (146) may include a history of page clicks and/or text inputs performed by the user to track the user's interaction with the software application. A user activity may, thus, be documented by storing an identifier for the user activity in the clickstream. In combination, user activity gathered over time may establish a context that may help identify a problem that the user is experiencing. The level of detail of user activity documented in the clickstream may vary. While in some scenarios, the clickstream may document all or almost all user activity, in other scenarios, not all user activities may be documented. For example, privacy requirements may exclude text typed by the user or other sensitive data from the clickstream. Further, the granularity of the clickstream may vary. In some scenarios, each user activity may be documented, whereas in other scenarios, only summaries of user activities may be documented. For example, counts of clicks may be stored in the clickstream rather than the individual clicks. In some embodiments, page or screen identifiers (IDs) for pages or screens that the user has accessed may be documented in the clickstream. Additional information may be included. For example, the time spent on a particular screen or page, interactions of the software application with third party components (such as when importing or downloading (successfully or unsuccessfully) external data such as bank account information, forms, etc.), may be included as well.” See also Podgorny ¶0060, “To consider the sequence (or order) of the elements in the clickstream, a recurrent neural network (RNN) may be used. The RNN accepts, at the input, a sequence of vectors encoding the clickstream to produce a sequence of vectors representing hidden layer outputs. These hidden layer output vectors may subsequently be processed by an output layer which may implement, for example, a softmax function.” Podgorny discloses the concept of tracking a sequence of interactions performed by the consumer including online interactions while navigating different pages.)
Although Podgorny discloses the above-enclosed invention, Podgorny fails to explicitly disclose the concept of further tracking failed attempts.
However Cheng as shown, which talks about solution keyword tagging, teaches the concept of tracking failed attempts.
(See Cheng ¶0006, “In response to the computer receiving an indication that a tried solution in the solution keyword tag cloud did not resolve the issue experienced by the user, the computer updates the solution keyword tag cloud by moving the tried solution that failed to resolve the issue from a solution section of the solution keyword tag cloud to a condition section of the solution keyword tag cloud and updates the solution context-clearness index based on the tried solution failing to resolve the issue. By updating the solution keyword tag cloud and solution context-clearness index as solutions are tried without resolving the issue, the computer continuously organizes tried and untried solutions for the user, lets the user know how far the user has come in resolving the issue, and informs the user as to how close the user is to finding the correct technical solution to the issue.” Cheng teaches the concept of tracking failed attempts for solutions.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Cheng with the invention of Podgorny. As shown Podgorny discloses the concept of performing analysis on user inputs including pages visited and the sequence of inputs for determining issues and solutions for the user. Cheng further teaches the concept of tracking failed solutions associated with keywords/problems. Cheng teaches this concept to further map and update issues and solutions including removing particular associations which failed to further quickly and directly find the appropriate solutions (See Cheng ¶0006). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Cheng to further refine the issue-solution identification based on the both positive and negative feedback.
provide the tracked customer attributes, including the order and type of pages visited and the failed attempts, while navigating online in the browsing sequence to a predictive machine learning model, to determine during the browsing sequence, a prediction of a primary intent for seeking assistance comprising: (See Podgorny ¶0022, “Turning to FIG. 1, a system (100), in accordance with one or more embodiments of the disclosure, is shown. On a high level, the system (100) employs natural language processing to generate dynamic agent notes in the form of a problem summary (162). The problem summary (162) may be generated in real-time, as a support agent (198) is interacting with a user (196). The generation of a problem summary may be performed in real-time, as the support agent (198) and the user (196) are interacting, e.g., in a conversation.” See also Podgorny ¶0047, “The flowchart of FIG. 4 describes a method for decoding a problem that the user is experiencing from multimodal information in accordance with one or more embodiments of the disclosure. The method relies on multimodal input (including an interaction between a user and a support agent, a clickstream, and optionally user attributes) to provide a problem summary. The method may be executed whenever an interaction between the user and the support agent becomes available. The method may be re-executed as the interaction continues. For example, the method may be initially executed when the user asks a question or reports a problem. The method may be re-executed when the user provides additional details, e.g., in response to the support agent asking for clarification. With a more comprehensive interaction between the user and the support agent becoming available over time, the problem summary that is generated by the method may become increasingly detailed and/or accurate by more closely reflecting the actual problem that the user is experiencing. Additional methods may be executed in conjunction with the method of FIG. 4. For example, to update machine learning-based components used in the method of FIG. 4, the steps described in FIG. 6 may be executed, as discussed below.” Podgorny discloses the concept of performing real time tracking and analysis of interactions for determining the issue the end-user is seeking a solution for.)
at least one predicted problem encountered by the customer associated with the tracked customer attributes and a context of actions derived from the customer attributes, (See Podgorny ¶0040, “The shared latent space decoder (240), in one or more embodiments, operates on the shared latent space representation (230) to produce the problem summary (242). The problem summary may be a human-readable text, for example, a question summarizing the user's problem based on the information that was obtained from the interactions between the user and the support agent (204), the clickstream (206), and optionally the user attributes (208). The shared latent space decoder (240) may be an artificial neural network such as a long short-term memory (LSTM) model, discussed in detail below. Various hyperparameters may be used to tune the LSTM model for a given dataset. The hyperparameters may include, but are not limited to, the number of layers and the dimensionality of the hidden state.” Podgorny discloses the concept of providing collected information to a learning machine to determine a consumer problem.)
the model trained based on prior historical behaviour of other customers in the entity comprising browser navigational flows for others indicative of a known associated problem; (See Podgorny ¶0058, “The individual entries in the clickstream are categorical variables such as screen IDs and may, thus, be treated analogous to text (letter, tokens, words, etc.). Consider the previously introduced example of the clickstream [“incomeexplore”, “s1040perbfdi8858”, “2017 deductionscredits”, “deductionsguideme”] where the screen IDs form a sequence of categorical variables. These screen IDs or other elements of the clickstream may be processed using, for example, an artificial neural network to generate a clickstream embedding. Historically observed sequences of screen IDs obtained from click streams gathered over time may form the corpus used for training the artificial neural network.” Podgorny discloses the concept of the model being trained based on historical interaction information including the clickstream information.)
dynamically determine a solution to the predicted problem, while browsing in the browsing sequence, using the predictive machine learning model based on accessing a database linking similar problems and associated solutions; and (See Podgorny ¶0077, “FIG. 6 discusses methods for training various machine learning-based components of the system. Embodiments of the disclosure employ generative models that are based on adversarial training to produce a problem summary, and subsequently to identify a suggested solution. The models “translate” from the clickstream obtained from a user and interactions between the user and a support agent. Some of the training may be performed as an updating after one or more executions of the method of FIG. 4 once feedback from the user and/or the support agent becomes available to guide the training, as discussed below.” See also Podgorny ¶0020, “Embodiments of the disclosure may enable the generation of the problem summary in real time. A suggested solution may also be provided by matching the problem summary to that of a closest previously solved case.” Podgorny discloses determining a solution to the determined problem based on historical information problem and solutions using machine learning.)
Although the combination of Podgorny and Cheng discloses the above-enclosed invention, the combination fails to explicitly disclose the concept of providing the solution and context to the consumer.
However Douglas as shown, which talks about customer service prediction, teaches the concept of providing the solution and context to the consumer through a user interface.
present the solution and associated context of the solution to a user interface of the computer device associated with the computer application for the customer; (See Douglas ¶0058, “FIG. 7 is an exemplary customer interface 700. Customer interface 700 may be displayed on a mobile device, kiosk, or ATM. Customer interface 700 may require a customer to sign into their account by providing details such as a password, personal identification number, or credit/debit card. Customer interface 700 may display the customer name and account number 702 associated with the customer. Customer interface 700 may display a prompt 704, based on detected customer activity via an app, website, or directly at a kiosk, asking if the customer is trying to complete a specific action, e.g., make a transfer, make a deposit, check an account balance, etc. Customer interface 700 may also display a help button 706 to connect the customer with a customer service representative or to receive further input from the customer regarding their request.” Douglas teaches the concept of providing both a solution and a context to a consumer on an interface.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of the invention to have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Douglas with the combination of Podgorny and Cheng. As shown, the combination discloses the concept of performing analysis on consumer interaction to determine the consumer’s issue and identify potential solutions to improve consumer experience through a customer service agent. Douglas further teaches the concept of providing this information to both the customer service agent or directly to the consumer (See Douglas ¶0056). Douglas teaches this concept to further improve customer service experience for the consumers by improving agent knowledge and reducing resolution times by providing the solution for the consumer to execute (See Douglas ¶0001-¶0003). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized teachings of Douglas as Douglas teaches providing the context information and solution to either the agent or directly to the consumer, thereby reducing the time required to resolve the issue by either minimizing the time for an agent interaction or eliminating the need for agent interaction.
Although the combination of Podgorny, Cheng, and Douglas discloses the above-enclosed invention including the concept of utilizing feedback to further train the machine learning model., the combination fails to explicitly disclose the concept of prompting feedback directly from the user/consumer.
However Jungmeisteris as shown, which talks about real time analysis of customer sentiment for automated customer service, teaches the concept of prompting feedback directly from the user/consumer.
prompt for feedback on the user interface of the computer device in response to the solution presented and (See Jungmeisteris ¶0062, “An example of one such interface is illustrated with reference to FIGS. 4A and 4B. FIGS. 4A and 4B each depict a user interface (400 and 410, respectively) which inquire whether the user's problem was solved. These UIs may be generally understood as intercept surveys, that is, surveys presented during the user's workflow or progression through the website or application. In FIG. 4A, the user's problem was resolved, and a progression of screens is displayed in which system 110 requests additional information from the user (402), takes in additional input from the user (404), requests free-form text input with user feedback (406), and ends the interaction (408). In FIG. 4B, the user's problem was not resolved, and a different progression of screens is displayed, though the general process of requesting (412) and receiving (414) information, and requesting freeform input (416) remains. On UI 418, rather than ending the interaction, system 110 displays options for self-solve actions 418-1, 418-2, and 418-3, as well as a button 418-4 to facilitate escalation from a self-solve solution to an agent-implemented solution.” Jungmeisteris teaches the concept of prompting the consumer for feedback regarding the problem and solution.)
track the feedback to revise the training of the model based on the feedback. (See Podgorny ¶0087, “Further, based on the summery of the user's problem, a relevant answer may be obtained and provided to the support agent while the support agent is interacting with the user. With a large volume of previously solved cases, the likeliness of a newly received support request being similar to a previously processed support request increases. Accordingly, embodiments of the disclosure may learn to provide accurate answers, based on previously handled interactions between users and support agents. The recommendation provided by the support agent may thus be better informed, based on relevant answers being proposed by embodiments of the disclosure.” Podgorny teaches the concept of utilizing the feedback to further train the model.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Jungmeisteris with the combination of Podgorny, Cheng, and Douglas. As shown, the combination discloses the concept of utilizing feedback to further train the learning model for both problem identification as well as solutions. Jungmeisteris further teaches the concept of soliciting consumer feedback during the interaction including directly after providing solutions. Jungmeisteris teaches this concept to allow for accurate and timely collection of feedback and sentiment analysis and addressing issues with existing survey feedbacks during support interactions (See Jungmeisteris ¶0003-¶0005). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Jungmeisteris to further improve the collection of timely and accurate feedback over exiting survey based feedback collection.
As per claim 3:
The system of claim 1 wherein tracking the customer attributes of the online customer behaviour further comprises the instructions configuring the processor to track flow of user events on the computer application including browsing to navigate to one of:
select online assistance using the application;
browse to an informational web page for reviewing frequently asked questions;
browse to a support web page for obtaining assistance; and
initiate a chat session to request assistance from a support resource.
(See Podgorny ¶0052, “In one embodiment, a series of screen IDs are collected as the clickstream when the user navigates through the software application, thereby accessing a series of screens or pages of the software application. The following example is based on four screens of a tax software application being sequentially accessed. The screen IDs are stored in the array with screen IDs such as [“incomeexplore”, “s1040perbfdi8858”, “2017deductionscredits”, “deductionsguideme”]. The array may have a set size and may be limited to, for example, the 4, 5, or 20 most recent screen IDs. “Null” values may be entered initially, before the user begins accessing the software application, and these null values may be replaced by actual screen IDs as the user is accessing screens of the software application. The collected screen IDs forming the clickstream are categorical variables and may, thus, directly serve as the input to the query decoder (150). Alternatively, the screen IDs may be transformed into another representation using any kind of alphanumerical encoding to identify the screens that were accessed by the user.” See also Podgorny ¶0042, “The method relies on multimodal input (including an interaction between a user and a support agent, a clickstream, and optionally user attributes) to provide a problem summary. The method may be executed whenever an interaction between the user and the support agent becomes available.” Podgorny discloses the concept of tracking consumer behavior to include tracking user browsing activity including accessing support services.)
As per claim 5:
The system of claim 1, wherein the customer attributes are selected from the group consisting of: customer interaction behaviour online, customer interactions with the computer application, location of the computer device during the navigational events, time frame of the navigational events, length of each interaction in the navigational events, and particular sequence of interactions in the navigational events with at least one of a website and the customer application leading to a request for assistance. (See Podgorny ¶0030, “The clickstream (146) may be processed, for example, by performing a statistical analysis of the clickstream. The statistical analysis may provide insights into the user's behavior and/or interest. For example, a high number of repetitive clicks and/or significant time spent on a particular page may imply that the user is experiencing difficulties on this page. The clickstream (146) may, thus, provide context for the identification of the user's problem. The obtaining of the clickstream is described below with reference to the flowchart of FIG. 4.” Podgorny discloses the concept of the user attribute to include user behavior.)
As per claim 6:
The system of claim 5, wherein the customer attributes further define the context of actions, and the context of actions is further used to refine the predicted intent based on other users having a similar context of actions and the navigational events to the particular customer. (See Podgorny ¶0067, “In Step 410, the problem summary is decoded from the shared latent space representation. The decoding may be performed by an LSTM, similar to the previously introduced LSTM, trained or updated as described in Step 604 of FIG. 6, to obtain a problem summary that is human-readable text, for example, a question summarizing the user's problem. The decoding is based on a supervised learning approach that uses historical data (e.g., the previously experienced and documented support calls). The shared latent space representation leverages the additional information that becomes available through the clickstream. As a result, the decoding result may be considerably better than what would be achievable by decoding from the question alone. An example technique for decoding using LSTM is described in U.S. patent application Ser. No. 15/994,898, which is incorporated herein by reference.” Podgorny discloses the concept of utilizing consumer attributes to determine context and intent, including determining intent based on previous similar context and attributes.)
As per claim 7:
The system of claim 1, wherein the model utilizes the particular flow of navigational events for the customer leading to the request for assistance online on the computer application for prediction of intent based on determining a similarity of the particular flow of navigational events to prior similar customer navigational events leading to a defined request for assistance for other customers interacting with the application. (See Podgorny ¶0038, “The clickstream encoder (220), in one or more embodiments, processes the clickstream (206) to output the clickstream embedding (216). Various algorithms may be used to obtain the clickstream embedding (216) from the clickstream (206). In one or more embodiments, a deep learning-type algorithm is used. The algorithm may be, for example, a convolutional neural network (CNN). The CNN may include convolutional layers, pooling layers and fully connected layers. The CNN may accept the elements of the clickstream as input, and may provide a classification of the clickstream, based on a training or updating of the CNN. This training may have been performed using reinforcement learning, as discussed in detail with reference to FIG. 7. In one embodiment, a recurrent neural network (RNN) may be used. The RNN accepts, at the input, a sequence of vectors encoding the elements of the clickstream to produce a sequence of vectors representing hidden layer outputs. These hidden layer output vectors may subsequently be processed by an output layer which may implement, for example, a softmax function. In one or more embodiments, a Long Short-Term Memory (LSTM) type RNN is used, as discussed in detail below.” Podgorny discloses the concept of the learning model utilizing the click/navigation stream and historical navigation information to predict the issue.)
As per claim 8:
The system of claim 7, wherein the model further utilizes the particular flow of navigational events to retrieve associated known problems encountered by the other customers to automatically predict the one or more problems likely encountered by the customer. (See Podgorny ¶0073, “In Step 502, the linguistic segments that represent the user problem are identified. Additionally, other classes of linguistic segments may also be identified. For example, linguistic segments that include a verification of the user problem by the support agent, linguistic segments that include a proposal of solution or action by the support agent, and/or linguistic segments that include a user validation of the support agent's proposed solution or action may be identified. The detection may be repeatedly performed during the ongoing interaction between the user and the support agent.” Podgorny discloses the concept of utilizing the clickstream information to determine the consumer problem.)
As per claim 10:
A computer implemented method for dynamically providing predictive context-aware solutions on computing devices to online customers, the method comprising:
tracking, by a processor, customer attributes comprising: an online customer behaviour of a particular customer of an entity when interacting with a computer application associated with the entity and on a computer device, the online customer behaviour tracked comprising: a particular flow of navigational events when browsing the computer application indicative of the particular customer seeking assistance, the flow (See Podgorny ¶0036, “Turning to FIG. 2, a problem decoder (200), in accordance with one or more embodiments, is shown. The problem decoder (200) includes a user input encoder (210), a clickstream encoder (220), a shared latent space representation (230), and a shared latent space decoder (240). In combination, these components may output a problem summary (242) from the modalities (202), provided as inputs. The modalities may include the interaction (204) between the user and the support agent, the clickstream (206), and optionally user attributes (208). Other additional components, for example, as shown in FIG. 7, may exist and may be used to update components of the problem decoder (200) using a reinforcement learning paradigm.” Podgorny discloses tracking consumer behavior including navigation information through an application.)
including a sequence of interactions online, order and type of previous pages visited and failed attempts while navigating online in a browsing sequence; (See Podgorny ¶0029, “In one or more embodiments, the application back-end (140) receives a user input provided by the user via the input interface (122) and generates a clickstream (146) from the user input. Broadly speaking, the clickstream (146) may be generated by the application back-end (140) as the user is navigating through the software application. The clickstream (146) may document any type of interaction of the user with the software application. For example, the clickstream (146) may include a history of page clicks and/or text inputs performed by the user to track the user's interaction with the software application. A user activity may, thus, be documented by storing an identifier for the user activity in the clickstream. In combination, user activity gathered over time may establish a context that may help identify a problem that the user is experiencing. The level of detail of user activity documented in the clickstream may vary. While in some scenarios, the clickstream may document all or almost all user activity, in other scenarios, not all user activities may be documented. For example, privacy requirements may exclude text typed by the user or other sensitive data from the clickstream. Further, the granularity of the clickstream may vary. In some scenarios, each user activity may be documented, whereas in other scenarios, only summaries of user activities may be documented. For example, counts of clicks may be stored in the clickstream rather than the individual clicks. In some embodiments, page or screen identifiers (IDs) for pages or screens that the user has accessed may be documented in the clickstream. Additional information may be included. For example, the time spent on a particular screen or page, interactions of the software application with third party components (such as when importing or downloading (successfully or unsuccessfully) external data such as bank account information, forms, etc.), may be included as well.” See also Podgorny ¶0060, “To consider the sequence (or order) of the elements in the clickstream, a recurrent neural network (RNN) may be used. The RNN accepts, at the input, a sequence of vectors encoding the clickstream to produce a sequence of vectors representing hidden layer outputs. These hidden layer output vectors may subsequently be processed by an output layer which may implement, for example, a softmax function.” Podgorny discloses the concept of tracking a sequence of interactions performed by the consumer including online interactions while navigating different pages.)
Although Podgorny discloses the above-enclosed invention, Podgorny fails to explicitly disclose the concept of further tracking failed attempts.
However Cheng as shown, which talks about solution keyword tagging, teaches the concept of tracking failed attempts.
(See Cheng ¶0006, “In response to the computer receiving an indication that a tried solution in the solution keyword tag cloud did not resolve the issue experienced by the user, the computer updates the solution keyword tag cloud by moving the tried solution that failed to resolve the issue from a solution section of the solution keyword tag cloud to a condition section of the solution keyword tag cloud and updates the solution context-clearness index based on the tried solution failing to resolve the issue. By updating the solution keyword tag cloud and solution context-clearness index as solutions are tried without resolving the issue, the computer continuously organizes tried and untried solutions for the user, lets the user know how far the user has come in resolving the issue, and informs the user as to how close the user is to finding the correct technical solution to the issue.” Cheng teaches the concept of tracking failed attempts for solutions.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Cheng with the invention of Podgorny. As shown Podgorny discloses the concept of performing analysis on user inputs including pages visited and the sequence of inputs for determining issues and solutions for the user. Cheng further teaches the concept of tracking failed solutions associated with keywords/problems. Cheng teaches this concept to further map and update issues and solutions including removing particular associations which failed to further quickly and directly find the appropriate solutions (See Cheng ¶0006). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Cheng to further refine the issue-solution identification based on the both positive and negative feedback.
providing, by the processor, the tracked customer attributes, including the order and type of pages visited and the failed attempts, while navigating online in the browsing sequence to a predictive machine learning model, to determine during the browsing sequence, a prediction of a primary intent for seeking assistance comprising: (See Podgorny ¶0022, “Turning to FIG. 1, a system (100), in accordance with one or more embodiments of the disclosure, is shown. On a high level, the system (100) employs natural language processing to generate dynamic agent notes in the form of a problem summary (162). The problem summary (162) may be generated in real-time, as a support agent (198) is interacting with a user (196). The generation of a problem summary may be performed in real-time, as the support agent (198) and the user (196) are interacting, e.g., in a conversation.” See also Podgorny ¶0047, “The flowchart of FIG. 4 describes a method for decoding a problem that the user is experiencing from multimodal information in accordance with one or more embodiments of the disclosure. The method relies on multimodal input (including an interaction between a user and a support agent, a clickstream, and optionally user attributes) to provide a problem summary. The method may be executed whenever an interaction between the user and the support agent becomes available. The method may be re-executed as the interaction continues. For example, the method may be initially executed when the user asks a question or reports a problem. The method may be re-executed when the user provides additional details, e.g., in response to the support agent asking for clarification. With a more comprehensive interaction between the user and the support agent becoming available over time, the problem summary that is generated by the method may become increasingly detailed and/or accurate by more closely reflecting the actual problem that the user is experiencing. Additional methods may be executed in conjunction with the method of FIG. 4. For example, to update machine learning-based components used in the method of FIG. 4, the steps described in FIG. 6 may be executed, as discussed below.” Podgorny discloses the concept of performing real time tracking and analysis of interactions for determining the issue the end-user is seeking a solution for.)
at least one predicted problem encountered by the customer associated with the tracked customer attributes and a context of actions derived from the customer attributes, (See Podgorny ¶0040, “The shared latent space decoder (240), in one or more embodiments, operates on the shared latent space representation (230) to produce the problem summary (242). The problem summary may be a human-readable text, for example, a question summarizing the user's problem based on the information that was obtained from the interactions between the user and the support agent (204), the clickstream (206), and optionally the user attributes (208). The shared latent space decoder (240) may be an artificial neural network such as a long short-term memory (LSTM) model, discussed in detail below. Various hyperparameters may be used to tune the LSTM model for a given dataset. The hyperparameters may include, but are not limited to, the number of layers and the dimensionality of the hidden state.” Podgorny discloses the concept of providing collected information to a learning machine to determine a consumer problem.)
the model being trained based on prior historical behaviour of other customers in the entity comprising browser navigational flows for others indicative of a known associated problem; (See Podgorny ¶0058, “The individual entries in the clickstream are categorical variables such as screen IDs and may, thus, be treated analogous to text (letter, tokens, words, etc.). Consider the previously introduced example of the clickstream [“incomeexplore”, “s1040perbfdi8858”, “2017 deductionscredits”, “deductionsguideme”] where the screen IDs form a sequence of categorical variables. These screen IDs or other elements of the clickstream may be processed using, for example, an artificial neural network to generate a clickstream embedding. Historically observed sequences of screen IDs obtained from click streams gathered over time may form the corpus used for training the artificial neural network.” Podgorny discloses the concept of the model being trained based on historical interaction information including the clickstream information.)
dynamically determining, by the processor, a solution to the predicted problem, while browsing in the browsing sequence, using the predictive machine learning model based on accessing a database linking similar problems and associated solutions; (See Podgorny ¶0077, “FIG. 6 discusses methods for training various machine learning-based components of the system. Embodiments of the disclosure employ generative models that are based on adversarial training to produce a problem summary, and subsequently to identify a suggested solution. The models “translate” from the clickstream obtained from a user and interactions between the user and a support agent. Some of the training may be performed as an updating after one or more executions of the method of FIG. 4 once feedback from the user and/or the support agent becomes available to guide the training, as discussed below.” See also Podgorny ¶0020, “Embodiments of the disclosure may enable the generation of the problem summary in real time. A suggested solution may also be provided by matching the problem summary to that of a closest previously solved case.” Podgorny discloses determining a solution to the determined problem based on historical information problem and solutions using machine learning.)
Although the combination of Podgorny and Cheng discloses the above-enclosed invention, the combination fails to explicitly disclose the concept of providing the solution and context to the consumer.
However Douglas as shown, which talks about customer service prediction, teaches the concept of providing the solution and context to the consumer through a user interface.
presenting, by the processor, the solution and associated context of the solution to a user interface of the computer device associated with the computer application for the customer; (See Douglas ¶0058, “FIG. 7 is an exemplary customer interface 700. Customer interface 700 may be displayed on a mobile device, kiosk, or ATM. Customer interface 700 may require a customer to sign into their account by providing details such as a password, personal identification number, or credit/debit card. Customer interface 700 may display the customer name and account number 702 associated with the customer. Customer interface 700 may display a prompt 704, based on detected customer activity via an app, website, or directly at a kiosk, asking if the customer is trying to complete a specific action, e.g., make a transfer, make a deposit, check an account balance, etc. Customer interface 700 may also display a help button 706 to connect the customer with a customer service representative or to receive further input from the customer regarding their request.” Douglas teaches the concept of providing both a solution and a context to a consumer on an interface.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of the invention to have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Douglas with the combination of Podgorny and Cheng. As shown, the combination discloses the concept of performing analysis on consumer interaction to determine the consumer’s issue and identify potential solutions to improve consumer experience through a customer service agent. Douglas further teaches the concept of providing this information to both the customer service agent or directly to the consumer (See Douglas ¶0056). Douglas teaches this concept to further improve customer service experience for the consumers by improving agent knowledge and reducing resolution times by providing the solution for the consumer to execute (See Douglas ¶0001-¶0003). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized teachings of Douglas as Douglas teaches providing the context information and solution to either the agent or directly to the consumer, thereby reducing the time required to resolve the issue by either minimizing the time for an agent interaction or eliminating the need for agent interaction.
Although the combination of Podgorny, Cheng, and Douglas discloses the above-enclosed invention including the concept of utilizing feedback to further train the machine learning model., the combination fails to explicitly disclose the concept of prompting feedback directly from the user/consumer.
However Jungmeisteris as shown, which talks about real time analysis of customer sentiment for automated customer service, teaches the concept of prompting feedback directly from the user/consumer.
prompting, by the processor, for feedback on the user interface of the computer device in response to the solution presented and (See Jungmeisteris ¶0062, “An example of one such interface is illustrated with reference to FIGS. 4A and 4B. FIGS. 4A and 4B each depict a user interface (400 and 410, respectively) which inquire whether the user's problem was solved. These UIs may be generally understood as intercept surveys, that is, surveys presented during the user's workflow or progression through the website or application. In FIG. 4A, the user's problem was resolved, and a progression of screens is displayed in which system 110 requests additional information from the user (402), takes in additional input from the user (404), requests free-form text input with user feedback (406), and ends the interaction (408). In FIG. 4B, the user's problem was not resolved, and a different progression of screens is displayed, though the general process of requesting (412) and receiving (414) information, and requesting freeform input (416) remains. On UI 418, rather than ending the interaction, system 110 displays options for self-solve actions 418-1, 418-2, and 418-3, as well as a button 418-4 to facilitate escalation from a self-solve solution to an agent-implemented solution.” Jungmeisteris teaches the concept of prompting the consumer for feedback regarding the problem and solution.)
tracking, by the processor, the feedback to revise the training of the model based on the feedback. (See Podgorny ¶0087, “Further, based on the summery of the user's problem, a relevant answer may be obtained and provided to the support agent while the support agent is interacting with the user. With a large volume of previously solved cases, the likeliness of a newly received support request being similar to a previously processed support request increases. Accordingly, embodiments of the disclosure may learn to provide accurate answers, based on previously handled interactions between users and support agents. The recommendation provided by the support agent may thus be better informed, based on relevant answers being proposed by embodiments of the disclosure.” Podgorny teaches the concept of utilizing the feedback to further train the model.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Jungmeisteris with the combination of Podgorny, Cheng, and Douglas. As shown, the combination discloses the concept of utilizing feedback to further train the learning model for both problem identification as well as solutions. Jungmeisteris further teaches the concept of soliciting consumer feedback during the interaction including directly after providing solutions. Jungmeisteris teaches this concept to allow for accurate and timely collection of feedback and sentiment analysis and addressing issues with existing survey feedbacks during support interactions (See Jungmeisteris ¶0003-¶0005). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Jungmeisteris to further improve the collection of timely and accurate feedback over exiting survey based feedback collection.
As per claim 12:
The method of claim 10 wherein tracking the customer attributes of the online customer behaviour further comprises tracking flow of user events on the computer application including browsing to navigate to one of: select online assistance using the application; browsing to an informational web page for reviewing frequently asked questions; browse to a support web page for obtaining assistance; and initiate a chat session to request assistance from a support resource. (See Podgorny ¶0052, “In one embodiment, a series of screen IDs are collected as the clickstream when the user navigates through the software application, thereby accessing a series of screens or pages of the software application. The following example is based on four screens of a tax software application being sequentially accessed. The screen IDs are stored in the array with screen IDs such as [“incomeexplore”, “s1040perbfdi8858”, “2017deductionscredits”, “deductionsguideme”]. The array may have a set size and may be limited to, for example, the 4, 5, or 20 most recent screen IDs. “Null” values may be entered initially, before the user begins accessing the software application, and these null values may be replaced by actual screen IDs as the user is accessing screens of the software application. The collected screen IDs forming the clickstream are categorical variables and may, thus, directly serve as the input to the query decoder (150). Alternatively, the screen IDs may be transformed into another representation using any kind of alphanumerical encoding to identify the screens that were accessed by the user.” See also Podgorny ¶0042, “The method relies on multimodal input (including an interaction between a user and a support agent, a clickstream, and optionally user attributes) to provide a problem summary. The method may be executed whenever an interaction between the user and the support agent becomes available.” Podgorny discloses the concept of tracking consumer behavior to include tracking user browsing activity including accessing support services.)
As per claim 14:
The method of claim 10, wherein the customer attributes are selected from the group consisting of: customer interaction behaviour online, customer interactions with the computer application, location of the computer device during the navigational events, time frame of the navigational events, length of each interaction in the navigational events, and particular sequence of interactions in the navigational events with at least one of a website and the customer application leading to a request for assistance. (See Podgorny ¶0030, “The clickstream (146) may be processed, for example, by performing a statistical analysis of the clickstream. The statistical analysis may provide insights into the user's behavior and/or interest. For example, a high number of repetitive clicks and/or significant time spent on a particular page may imply that the user is experiencing difficulties on this page. The clickstream (146) may, thus, provide context for the identification of the user's problem. The obtaining of the clickstream is described below with reference to the flowchart of FIG. 4.” Podgorny discloses the concept of the user attribute to include user behavior.)
As per claim 15:
The method of claim 14, wherein the customer attributes further define the context of actions, and the context of actions is further used to refine the predicted intent based on other users having a similar context of actions and the navigational events to the particular customer. (See Podgorny ¶0067, “In Step 410, the problem summary is decoded from the shared latent space representation. The decoding may be performed by an LSTM, similar to the previously introduced LSTM, trained or updated as described in Step 604 of FIG. 6, to obtain a problem summary that is human-readable text, for example, a question summarizing the user's problem. The decoding is based on a supervised learning approach that uses historical data (e.g., the previously experienced and documented support calls). The shared latent space representation leverages the additional information that becomes available through the clickstream. As a result, the decoding result may be considerably better than what would be achievable by decoding from the question alone. An example technique for decoding using LSTM is described in U.S. patent application Ser. No. 15/994,898, which is incorporated herein by reference.” Podgorny discloses the concept of utilizing consumer attributes to determine context and intent, including determining intent based on previous similar context and attributes.)
As per claim 16:
The method of claim 10, wherein the model utilizes the particular flow of navigational events for the customer leading to the request for assistance online on the computer application for prediction of intent based on determining a similarity of the particular flow of navigational events to prior similar customer navigational events leading to a defined request for assistance for other customers interacting with the application. (See Podgorny ¶0038, “The clickstream encoder (220), in one or more embodiments, processes the clickstream (206) to output the clickstream embedding (216). Various algorithms may be used to obtain the clickstream embedding (216) from the clickstream (206). In one or more embodiments, a deep learning-type algorithm is used. The algorithm may be, for example, a convolutional neural network (CNN). The CNN may include convolutional layers, pooling layers and fully connected layers. The CNN may accept the elements of the clickstream as input, and may provide a classification of the clickstream, based on a training or updating of the CNN. This training may have been performed using reinforcement learning, as discussed in detail with reference to FIG. 7. In one embodiment, a recurrent neural network (RNN) may be used. The RNN accepts, at the input, a sequence of vectors encoding the elements of the clickstream to produce a sequence of vectors representing hidden layer outputs. These hidden layer output vectors may subsequently be processed by an output layer which may implement, for example, a softmax function. In one or more embodiments, a Long Short-Term Memory (LSTM) type RNN is used, as discussed in detail below.” Podgorny discloses the concept of the learning model utilizing the click/navigation stream and historical navigation information to predict the issue.)
As per claim 17:
The method of claim 16, wherein the model further utilizes the particular flow of navigational events to retrieve associated known problems encountered by the other customers to automatically predict the one or more problems likely encountered by the customer. (See Podgorny ¶0073, “In Step 502, the linguistic segments that represent the user problem are identified. Additionally, other classes of linguistic segments may also be identified. For example, linguistic segments that include a verification of the user problem by the support agent, linguistic segments that include a proposal of solution or action by the support agent, and/or linguistic segments that include a user validation of the support agent's proposed solution or action may be identified. The detection may be repeatedly performed during the ongoing interaction between the user and the support agent.” Podgorny discloses the concept of utilizing the clickstream information to determine the consumer problem.)
As per claim 19:
A non-transitory computer-readable medium containing computer program code that are executable by a processor for providing predictive context-aware solutions on computing devices to online customers, the processor to perform steps of:
tracking, by a processor, customer attributes comprising: an online customer behaviour of a particular customer of an entity when interacting with a computer application associated with the entity and on a computer device, the online customer behaviour tracked comprising: a particular flow of navigational events when browsing the computer application indicative of seeking assistance, the flow (See Podgorny ¶0036, “Turning to FIG. 2, a problem decoder (200), in accordance with one or more embodiments, is shown. The problem decoder (200) includes a user input encoder (210), a clickstream encoder (220), a shared latent space representation (230), and a shared latent space decoder (240). In combination, these components may output a problem summary (242) from the modalities (202), provided as inputs. The modalities may include the interaction (204) between the user and the support agent, the clickstream (206), and optionally user attributes (208). Other additional components, for example, as shown in FIG. 7, may exist and may be used to update components of the problem decoder (200) using a reinforcement learning paradigm.” Podgorny discloses tracking consumer behavior including navigation information through an application.)
including a sequence of interactions online, order and type of previous pages visited and failed attempts while navigating online in a browsing sequence; (See Podgorny ¶0029, “In one or more embodiments, the application back-end (140) receives a user input provided by the user via the input interface (122) and generates a clickstream (146) from the user input. Broadly speaking, the clickstream (146) may be generated by the application back-end (140) as the user is navigating through the software application. The clickstream (146) may document any type of interaction of the user with the software application. For example, the clickstream (146) may include a history of page clicks and/or text inputs performed by the user to track the user's interaction with the software application. A user activity may, thus, be documented by storing an identifier for the user activity in the clickstream. In combination, user activity gathered over time may establish a context that may help identify a problem that the user is experiencing. The level of detail of user activity documented in the clickstream may vary. While in some scenarios, the clickstream may document all or almost all user activity, in other scenarios, not all user activities may be documented. For example, privacy requirements may exclude text typed by the user or other sensitive data from the clickstream. Further, the granularity of the clickstream may vary. In some scenarios, each user activity may be documented, whereas in other scenarios, only summaries of user activities may be documented. For example, counts of clicks may be stored in the clickstream rather than the individual clicks. In some embodiments, page or screen identifiers (IDs) for pages or screens that the user has accessed may be documented in the clickstream. Additional information may be included. For example, the time spent on a particular screen or page, interactions of the software application with third party components (such as when importing or downloading (successfully or unsuccessfully) external data such as bank account information, forms, etc.), may be included as well.” See also Podgorny ¶0060, “To consider the sequence (or order) of the elements in the clickstream, a recurrent neural network (RNN) may be used. The RNN accepts, at the input, a sequence of vectors encoding the clickstream to produce a sequence of vectors representing hidden layer outputs. These hidden layer output vectors may subsequently be processed by an output layer which may implement, for example, a softmax function.” Podgorny discloses the concept of tracking a sequence of interactions performed by the consumer including online interactions while navigating different pages.)
. Although Podgorny discloses the above-enclosed invention, Podgorny fails to explicitly disclose the concept of further tracking failed attempts.
However Cheng as shown, which talks about solution keyword tagging, teaches the concept of tracking failed attempts.
(See Cheng ¶0006, “In response to the computer receiving an indication that a tried solution in the solution keyword tag cloud did not resolve the issue experienced by the user, the computer updates the solution keyword tag cloud by moving the tried solution that failed to resolve the issue from a solution section of the solution keyword tag cloud to a condition section of the solution keyword tag cloud and updates the solution context-clearness index based on the tried solution failing to resolve the issue. By updating the solution keyword tag cloud and solution context-clearness index as solutions are tried without resolving the issue, the computer continuously organizes tried and untried solutions for the user, lets the user know how far the user has come in resolving the issue, and informs the user as to how close the user is to finding the correct technical solution to the issue.” Cheng teaches the concept of tracking failed attempts for solutions.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Cheng with the invention of Podgorny. As shown Podgorny discloses the concept of performing analysis on user inputs including pages visited and the sequence of inputs for determining issues and solutions for the user. Cheng further teaches the concept of tracking failed solutions associated with keywords/problems. Cheng teaches this concept to further map and update issues and solutions including removing particular associations which failed to further quickly and directly find the appropriate solutions (See Cheng ¶0006). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Cheng to further refine the issue-solution identification based on the both positive and negative feedback.
providing, by the processor, the tracked customer attributes, including the order and type of pages visited and the failed attempts, while navigating online in the browsing sequence to a predictive machine learning model to determine a prediction during the browsing sequence, of a primary intent for seeking assistance comprising: (See Podgorny ¶0022, “Turning to FIG. 1, a system (100), in accordance with one or more embodiments of the disclosure, is shown. On a high level, the system (100) employs natural language processing to generate dynamic agent notes in the form of a problem summary (162). The problem summary (162) may be generated in real-time, as a support agent (198) is interacting with a user (196). The generation of a problem summary may be performed in real-time, as the support agent (198) and the user (196) are interacting, e.g., in a conversation.” See also Podgorny ¶0047, “The flowchart of FIG. 4 describes a method for decoding a problem that the user is experiencing from multimodal information in accordance with one or more embodiments of the disclosure. The method relies on multimodal input (including an interaction between a user and a support agent, a clickstream, and optionally user attributes) to provide a problem summary. The method may be executed whenever an interaction between the user and the support agent becomes available. The method may be re-executed as the interaction continues. For example, the method may be initially executed when the user asks a question or reports a problem. The method may be re-executed when the user provides additional details, e.g., in response to the support agent asking for clarification. With a more comprehensive interaction between the user and the support agent becoming available over time, the problem summary that is generated by the method may become increasingly detailed and/or accurate by more closely reflecting the actual problem that the user is experiencing. Additional methods may be executed in conjunction with the method of FIG. 4. For example, to update machine learning-based components used in the method of FIG. 4, the steps described in FIG. 6 may be executed, as discussed below.” Podgorny discloses the concept of performing real time tracking and analysis of interactions for determining the issue the end-user is seeking a solution for.)
at least one predicted problem encountered by the customer associated with the tracked customer attributes and a context of actions derived from the customer attributes, (See Podgorny ¶0040, “The shared latent space decoder (240), in one or more embodiments, operates on the shared latent space representation (230) to produce the problem summary (242). The problem summary may be a human-readable text, for example, a question summarizing the user's problem based on the information that was obtained from the interactions between the user and the support agent (204), the clickstream (206), and optionally the user attributes (208). The shared latent space decoder (240) may be an artificial neural network such as a long short-term memory (LSTM) model, discussed in detail below. Various hyperparameters may be used to tune the LSTM model for a given dataset. The hyperparameters may include, but are not limited to, the number of layers and the dimensionality of the hidden state.” Podgorny discloses the concept of providing collected information to a learning machine to determine a consumer problem.)
the model being trained based on prior historical behaviour of other customers in the entity comprising browser navigational flows for others indicative of a known associated problem; (See Podgorny ¶0058, “The individual entries in the clickstream are categorical variables such as screen IDs and may, thus, be treated analogous to text (letter, tokens, words, etc.). Consider the previously introduced example of the clickstream [“incomeexplore”, “s1040perbfdi8858”, “2017 deductionscredits”, “deductionsguideme”] where the screen IDs form a sequence of categorical variables. These screen IDs or other elements of the clickstream may be processed using, for example, an artificial neural network to generate a clickstream embedding. Historically observed sequences of screen IDs obtained from click streams gathered over time may form the corpus used for training the artificial neural network.” Podgorny discloses the concept of the model being trained based on historical interaction information including the clickstream information.)
dynamically determining, by the processor, a solution to the predicted problem, while browsing in the browsing sequence, using the predictive machine learning model based on accessing a database linking similar problems and associated solutions; (See Podgorny ¶0077, “FIG. 6 discusses methods for training various machine learning-based components of the system. Embodiments of the disclosure employ generative models that are based on adversarial training to produce a problem summary, and subsequently to identify a suggested solution. The models “translate” from the clickstream obtained from a user and interactions between the user and a support agent. Some of the training may be performed as an updating after one or more executions of the method of FIG. 4 once feedback from the user and/or the support agent becomes available to guide the training, as discussed below.” See also Podgorny ¶0020, “Embodiments of the disclosure may enable the generation of the problem summary in real time. A suggested solution may also be provided by matching the problem summary to that of a closest previously solved case.” Podgorny discloses determining a solution to the determined problem based on historical information problem and solutions using machine learning.)
Although the combination of Podgorny and Cheng discloses the above-enclosed invention, the combination fails to explicitly disclose the concept of providing the solution and context to the consumer.
However Douglas as shown, which talks about customer service prediction, teaches the concept of providing the solution and context to the consumer through a user interface.
presenting the solution, by the processor, and associated context of the solution to a user interface of the computer device associated with the computer application for the customer; (See Douglas ¶0058, “FIG. 7 is an exemplary customer interface 700. Customer interface 700 may be displayed on a mobile device, kiosk, or ATM. Customer interface 700 may require a customer to sign into their account by providing details such as a password, personal identification number, or credit/debit card. Customer interface 700 may display the customer name and account number 702 associated with the customer. Customer interface 700 may display a prompt 704, based on detected customer activity via an app, website, or directly at a kiosk, asking if the customer is trying to complete a specific action, e.g., make a transfer, make a deposit, check an account balance, etc. Customer interface 700 may also display a help button 706 to connect the customer with a customer service representative or to receive further input from the customer regarding their request.” Douglas teaches the concept of providing both a solution and a context to a consumer on an interface.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of the invention to have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Douglas with the combination of Podgorny and Cheng. As shown, the combination discloses the concept of performing analysis on consumer interaction to determine the consumer’s issue and identify potential solutions to improve consumer experience through a customer service agent. Douglas further teaches the concept of providing this information to both the customer service agent or directly to the consumer (See Douglas ¶0056). Douglas teaches this concept to further improve customer service experience for the consumers by improving agent knowledge and reducing resolution times by providing the solution for the consumer to execute (See Douglas ¶0001-¶0003). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized teachings of Douglas as Douglas teaches providing the context information and solution to either the agent or directly to the consumer, thereby reducing the time required to resolve the issue by either minimizing the time for an agent interaction or eliminating the need for agent interaction.
Although the combination of Podgorny, Cheng, and Douglas discloses the above-enclosed invention including the concept of utilizing feedback to further train the machine learning model., the combination fails to explicitly disclose the concept of prompting feedback directly from the user/consumer.
However Jungmeisteris as shown, which talks about real time analysis of customer sentiment for automated customer service, teaches the concept of prompting feedback directly from the user/consumer.
prompting, by the processor, for feedback on the user interface of the computer device in response to the solution presented and (See Jungmeisteris ¶0062, “An example of one such interface is illustrated with reference to FIGS. 4A and 4B. FIGS. 4A and 4B each depict a user interface (400 and 410, respectively) which inquire whether the user's problem was solved. These UIs may be generally understood as intercept surveys, that is, surveys presented during the user's workflow or progression through the website or application. In FIG. 4A, the user's problem was resolved, and a progression of screens is displayed in which system 110 requests additional information from the user (402), takes in additional input from the user (404), requests free-form text input with user feedback (406), and ends the interaction (408). In FIG. 4B, the user's problem was not resolved, and a different progression of screens is displayed, though the general process of requesting (412) and receiving (414) information, and requesting freeform input (416) remains. On UI 418, rather than ending the interaction, system 110 displays options for self-solve actions 418-1, 418-2, and 418-3, as well as a button 418-4 to facilitate escalation from a self-solve solution to an agent-implemented solution.” Jungmeisteris teaches the concept of prompting the consumer for feedback regarding the problem and solution.)
tracking, by the processor, feedback to revise the training of the model based on the feedback. (See Podgorny ¶0087, “Further, based on the summery of the user's problem, a relevant answer may be obtained and provided to the support agent while the support agent is interacting with the user. With a large volume of previously solved cases, the likeliness of a newly received support request being similar to a previously processed support request increases. Accordingly, embodiments of the disclosure may learn to provide accurate answers, based on previously handled interactions between users and support agents. The recommendation provided by the support agent may thus be better informed, based on relevant answers being proposed by embodiments of the disclosure.” Podgorny teaches the concept of utilizing the feedback to further train the model.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Jungmeisteris with the combination of Podgorny, Cheng, and Douglas. As shown, the combination discloses the concept of utilizing feedback to further train the learning model for both problem identification as well as solutions. Jungmeisteris further teaches the concept of soliciting consumer feedback during the interaction including directly after providing solutions. Jungmeisteris teaches this concept to allow for accurate and timely collection of feedback and sentiment analysis and addressing issues with existing survey feedbacks during support interactions (See Jungmeisteris ¶0003-¶0005). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Jungmeisteris to further improve the collection of timely and accurate feedback over exiting survey based feedback collection.
Claim(s) 9, 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Podgorny et al. (US 20210133581 A1) (hereafter Podgorny), in view of Cheng et al. (US 20180121929 A1) (hereafter Cheng), in view of Douglas (US 20200065825 A1) (hereafter Douglas), in view of Jungmeisteris et al. (US 20220374956 A1) (hereafter Jungmeisteris), in view of Ross et al. (US 20130051544 A1) (hereafter Ross).
As per claim 9:
Although the combination of Podgorny, Cheng, Douglas, and Jungmeisteris discloses the above-enclosed invention, the combination fails to explicitly disclose the concept of determining a consumer problem based on transaction information.
However Ross as shown, which talks about evaluation of consumer information for enhancing customer service interactions, teaches the concept of determining consumer problems based on transaction information.
The system of claim 8, wherein the instructions further configure the processor to:
utilize the tracked customer attributes, via the predictive machine learning model, comprising the particular flow of navigational events to initially predict an expected transaction to be performed at a future time subsequent to the navigational events based on a current sequence of interactive events; and
triggering a prediction by the predictive machine learning model of the intent and the at least one problem in response to determining that the expected transaction has not occurred at the future time, the predicting problem and the solution additionally based on the expected transaction.
(See Ross ¶0075, “For example, the customer 202 is contacting the call center in regard to a disputed credit card payment, the risk metric may indicate that the customer 202 always pays his/her credit card on time, thus there may be an issue with this payment. Subsequently a recommendation may be made to waive any late fee that may otherwise apply to the late payment.” Ross teaches the concept of tracking consumer interactions including transaction history, and predicting a problem of the consumer based on changes/difference in the transaction behavior.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of the invention to have utilized the teachings of Ross with the combination of Podgorny, Cheng, Douglas, and Jungmeisteris. As shown, the combination discloses the concept of tracking consumer interactions to predict consumer issues/problems prior to establishing communication with an agent. Ross further teaches this concept to include tracking transaction information and transaction history. Ross teaches this concept to further enhance service agent/representative capability by both pre-empting consumer issues as well as improving the consumer experience by providing beneficial alternatives or rewards for past good behaviors (See Ross ¶0006-¶0007). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Ross to further improve customer service interactions through tracking historical consumer behavior, and providing rewards and beneficial alternatives.
As per claim 18:
Although the combination of Podgorny, Cheng, Douglas, and Jungmeisteris discloses the above-enclosed invention, the combination fails to explicitly disclose the concept of determining a consumer problem based on transaction information.
However Ross as shown, which talks about evaluation of consumer information for enhancing customer service interactions, teaches the concept of determining consumer problems based on transaction information.
The method of claim 17, further comprising: the predictive machine learning model configured to
utilize the tracked customer attributes comprising the particular flow of navigational events to initially predict an expected transaction to be performed at a future time subsequent to the navigational events based on a current sequence of interactive events; and
triggering a prediction by the machine learning model of the intent and the at least one problem in response to determining that the expected transaction has not occurred at the future time, the predicting problem and the solution additionally based on the expected transaction.
(See Ross ¶0075, “For example, the customer 202 is contacting the call center in regard to a disputed credit card payment, the risk metric may indicate that the customer 202 always pays his/her credit card on time, thus there may be an issue with this payment. Subsequently a recommendation may be made to waive any late fee that may otherwise apply to the late payment.” Ross teaches the concept of tracking consumer interactions including transaction history, and predicting a problem of the consumer based on changes/difference in the transaction behavior.)
Therefore it would have been obvious to one of ordinary skill in the art at the time of the invention to have utilized the teachings of Ross with the combination of Podgorny, Cheng, Douglas, and Jungmeisteris. As shown, the combination discloses the concept of tracking consumer interactions to predict consumer issues/problems prior to establishing communication with an agent. Ross further teaches this concept to include tracking transaction information and transaction history. Ross teaches this concept to further enhance service agent/representative capability by both pre-empting consumer issues as well as improving the consumer experience by providing beneficial alternatives or rewards for past good behaviors (See Ross ¶0006-¶0007). Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have utilized the teachings of Ross to further improve customer service interactions through tracking historical consumer behavior, and providing rewards and beneficial alternatives.
Response to Arguments
Applicant's arguments filed 11/25/2025 have been fully considered but they are not persuasive.
In response to the Applicant’s arguments as directed towards the 35 U.S.C. 103 rejection under the combination of Podgorny, Cheng, Douglas, and Jungmeisteris, the Examiner respectfully disagrees. The Applicant’s Representative asserts the combination of Podgorny, Cheng, Douglas, and Jungmeisteris fail to teach/suggest the claimed invention, specifically the combination does not teach/suggest continuous, real-time, and in-session analysis and determination of consumer issues. The Examiner notes as further shown above, although Podgorny discloses the concept of performing analysis based on interaction between a user and support agent, Podgorny refers to the interaction as being part of the browsing session including tracking and updating the problem/solution analysis based on updated clickstream information. Still furthermore, Podgorny discloses the clickstream to also include a history of browsing activities including an order/sequence of browsing activity. In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). As shown, although Podgorny does not teach/suggest the concept of providing the problem/solution information to the user, Douglas does further teach this concept including the system automatically determining the problem/solution based on user interactions. Thus it would have been obvious to one of ordinary skill in the art at the time of filing to have modified the invention of Podgorny with the teachings of Douglas, whereby the interactions and information regarding user activities and interactions are collected and analyzed as disclosed by Podgorny, wherein the result of the analysis is further provided to the user. As such, the Examiner asserts the combination of Podgorny, Cheng, Douglas, and Jungmeisteris teaches the invention as currently claimed and rejection has been maintained.
All rejections made towards the dependent claims are maintained due to the lack of a reply by the applicant in regards to distinctly and specifically point out the supposed errors in the Examiner’s action in the prior Office Action (37 CFR 1.111). The Examiner asserts that the applicant only argues that the dependent claims should be allowable because the independent claims are unobvious and patentable over Podgorny, Cheng, Douglas, and Jungmeisteris, and where applicable, further in view of Ross. Therefore, the Examiner asserts that the combination of Podgorny, Cheng, Douglas, and Jungmeisteris teaches the invention as claimed and the rejection has been maintained.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Lin et al. (US 20220101148 A1), which talks about solution determination including tracking sequence of actions and results in troubleshooting.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to VINCENT M CAO whose telephone number is (571)270-5598. The examiner can normally be reached Monday - Friday 11-7.
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, ILANA SPAR can be reached at (571) 270-7537. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/VINCENT M CAO/Primary Examiner, Art Unit 3622