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 .
Status of Claims
The office action is being examined in response to the application filed by the applicant on December 15th, 2025.
Claims 1, 12, 16 and 20 have been amended and are hereby entered.
Claims 1 - 20 are pending and have been examined.
This action is made FINAL.
Response to Arguments
Applicant’s arguments filed on December 15h, 2025 have been fully considered but are not persuasive for the following reasons set forth below.
Regarding the applicant’s arguments against the 101 rejection of claims 1 - 20 on pages 12 – 16: Applicant argues that the claims are not directed to a mental process because a human cannot configure telecommunications links such as FDD, TDD, LTE, or millimeter wave links. This argument is not persuasive. The Examiner notes that the focus of the claims, under the broadest reasonable interpretation (BRI), is on: (a) receiving a customer communication, (b) creating a numerical representation (embedding), (c) comparing the embedding to stored embeddings, (d) identifying an intent classification, and (e) redirecting the communication based on that classification. All of these steps collectively describe collecting information, analyzing it, and making a decision, which falls squarely within: (i) mental processes (observation, evaluation, judgment), and (ii) certain methods of organizing human activity, specifically managing interactions between a customer and a service provider (e.g., routing customer requests). Such activities can be (and historically have been) performed by human operators (e.g., call center agents listening to a request and routing to the appropriate department).
Step 2A Prong 1: Applicant argues that the claims are not directed to a judicial exception, asserting that the claimed features cannot be performed in the human mind, particularly the limitation of: “redirecting, by configuring… a second millimeter wave link.” This argument is not persuasive. As stated above, under the BRI, the claims describe steps of collecting, analyzing, and acting on information, which falls within the abstract idea groupings of: mental processes and certain methods of organizing human activity, particularly managing interactions between a customer and a service provider.
Applicant’s reliance on the recitation of telecommunication features (e.g., FDD, TDD, LTE, millimeter wave links) is unpersuasive because these features do not alter the fundamental nature of the claim. The recited “configuring” of communication links is expressed at a high level of generality and in functional terms, without specifying how the network is technically modified or improved. As such, these elements are considered part of the environment or field of use in which the abstract idea is applied, and do not negate the recitation of a judicial exception—refer to MPEP § 2106.04(a)).
Step 2A Prong 2: Applicant argues that the claims integrate any alleged abstract idea into a practical application by improving telecommunications network operations, including reducing latency and improving routing efficiency. This argument is not persuasive.
The claims do not recite any specific improvement to the functioning of a computer or telecommunication network. Rather, the claims merely use generic computing components (e.g., server system, models, vector database), perform generic data processing operations (e.g., classification and comparison), as well as apply the result to route communications. The “redirecting… by configuring one or more communication links…” limitation is recited at a result-oriented level, without any technical detail describing how such configuration is performed or how it improves underlying network functionality. Further, there is no disclosure in the claims of specific network protocols being modified, improvements to signal transmission, bandwidth allocation, or latency at a technical level, or any particularized implementation of FDD, TDD, LTE or millimeter wave technologies. Instead, the claims merely use a telecommunications network as a tool to implement the abstract idea of classifying and routing customer communications. Accordingly, the claims do not integrate the abstract idea into a practical application (see MPEP §§ 2106.05(a). 2106.05(f), and 2106.05(g)).
Step 2B: Applicant argues that the claims recite significantly more than the alleged judicial exception, asserting that the features are not well-understood and routine, citing the Berkheimer memorandum. This argument is not persuasive. The additional elements recited in the claims include: (a) a server system, (b) machine learning models, (c) embedding vectors, (d) a vector database, and (e) communication links within a telecommunications network. These elements are described at a high level of generality and perform their expected functions, such as: processing data, generating numerical representations, comparing data as well as routing communications.
The claims do not recite any specific technical improvement or non-generic arrangement of these elements. Furthermore, the alleged “configuration” of communication links is not described with sufficient specificity to demonstrate that it is anything more than a generic implementation of routing within a network.
With respect to Applicant’s Berkheimer argument, the Examiner notes that the claims themselves do not recite any features that would indicate the presence of an unconventional technological improvement. When claims recite only generic components performing generic functions, no additional evidentiary showing is required—see MPEP § 2106.05(d)).
Considering the claim elements individually and as an ordered combination, the claims amount to no more than implementing the abstract idea using generic computing and networking components, which does not constitute significantly more.
For the reasons discussed above, Applicant’s arguments have been fully considered but are not persuasive. The claims remain directed to an abstract idea without integration into a practical application, and the additional elements do not amount to significantly more than the judicial exception.
Regarding the applicant’s arguments against the 1-3 rejection of claims 1 – 20 on pages 16 - 18: Applicant argues that Wohlwend does not disclose: “redirecting a phone call to a sub unit… including FDD, TDD, LTE, or millimeter wave links…” and asserts that the cited portions of Wohlwend merely describe transitioning between states in a graph, not configuring telecommunication links. This argument is not persuasive.
As previously set forth in the Office Action, Wohlwend discloses: (i) identifying user intent based on message input, and (ii) advancing processing to a different state/node in a graph based on the identified intent (see ¶¶ [0062] - [0064], FIGS. 5, 8B). These “states” correspond to different functional handling paths (e.g., BUY_TICKET, CHANGE_RESERVATION, CANCEL_RESERVATION), which are functionally equivalent to routing a communication to different service units. Thus, Wohlwend teaches: classification of user intent, and routing/redirecting the communication flow based on that intent. Under the broadest reasonable interpretation (BRI), “redirecting a phone call to a sub-unit” encompasses routing a communication session to a different processing path or service function, as performed in Wohlwend.
Thus, the rejection of claims 1 – 20 under 35 U.S.C. § 103 is maintained.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1 – 20 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claims 1, 12, and 16 have been amended to recite: “redirect… by configuring one or more communication links with the telecommunications network service provider…” and further recite that “the one or more configured communication links… or a millimeter wave link.” However, the originally filed specification does not provide adequate written description support for redirecting a communication session by configuring one or more communication links, nor for the recited specific communication link technologies (FDD, TDD, LTE, or millimeter wave) in the context of performing such redirection.
Applicant directs attention to paragraphs [¶0015], [¶0027], [¶0057], and [¶0067] as support for the amended limitations. These portions of the specification have been reviewed and are found to disclose only that: (a) a communication session may be redirected based on an identified intent; and (b) communication links may exist within a telecommunications network and may utilize technologies such as FDD, TDD, LTE, or millimeter wave. However, these paragraphs: (a) do not disclose or suggest that redirection is performed by configuring communication links, (b) do not disclose any mechanism, process, or structure for configuring communication links to achieve redirection, and (c) do not link the recited link technologies to the act of redirecting a communication session. Rather, the specification only broadly describes redirecting communications at a functional level without any indication that such redirection involves reconfiguration of communication links or manipulation of underlying network-layer operations.
For the reasons set forth above, claims 1 – 20 are rejected under 35 U.S.C. § 112(a) for failing to comply with the written description requirement. Claims 1, 12, and 16 are rejected for the reasons discussed above. Claims 2 – 11, 13 – 15, and 17 – 20 depend therefrom and, therefore, inherit the deficiencies of their respective parent claims, as they do not include additional limitations that would overcome the lack of written description.
Accordingly, claims 1 – 20 are rejected under 35 U.S.C. § 112(a).
Claim Rejections - 35 USC § 101
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.
Claims 1 – 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more, and therefore does not recite patent-eligible subject matter. Firstly, it should be stated that claim 1 is representative of the independent claim set 1, 12 and 16.
Step 2A Prong 1: The claims are directed to the abstract idea of collecting and analyzing data to make a classification or prediction, and then acting based on that prediction, which falls into both the category of mental processes (e.g., comparing, evaluating, classifying) and certain methods of organizing human activity, specifically managing interactions between people and engaging in commercial or legal interactions, such as routing customers to appropriate departments based on predicted needs. For instance, claim 1 recites:
receiving during a communication session, at a server system associated with the telecommunications network service provider, a communication from a customer, and
generating, by the server system, a first embedding vector representing a numerical representation of natural language extracted from the communication;
comparing, by the server system, the first embedding vector to multiple embedding vectors stored in a vector database associated with the server system,
identifying, based on the comparison, which intent classification of the multiple intent classifications is associated with the first embedding vector,
redirecting … the phone call to a sub-unit of the telecommunications network service provider based on the identified intent classification.
These limitations collectively describe an automated customer intent recognition and routing process—something that can be (and traditionally has been) performed by a human call center operator using intuition or a checklist. The claimed invention automates this process using numerical representations (embeddings) and a trained model, but these do not remove the concept from the realm of abstraction. These operations—extracting meaning from language, comparing it to known categories, and making a judgment—can all be performed by a human using mental steps or simple tools like pen and paper. The steps of receiving communications, comparing values, classifying them, and redirecting are examples of concepts performed in the human mind, including observation, evaluation, and judgment. These steps mirror the abstract idea group of “mental processes”, as outlined in MPEP 2106.04(a)(2)(III).
Step 2A Prong 2: For independent claims 1, 12 and 16, The judicial exception is not integrated into a practical application, because the claims as a whole, while looking for their additional element(s) of one or more processors; (from claim 16), a non-transitory memory (from claim 16), a server system (from claims 1 and 12) individually and in combination, merely is used as a tool to perform the abstract idea (refer to MPEP 2106.05(f)). Similarly, these limitations are “merely indicating a field of use or technological environment in which to apply a judicial exception do not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application” (MPEP 2106.05(h)). Therefore, this is indicative of the fact that the claim set has not integrated the abstract idea into a practical application and therefore, the claims are found to be directed to the abstract idea identified by the examiner.
Step 2B: For independent claims 1, 12 and 16, these claims recite the additional elements: one or more processors; (from claim 16), a non-transitory memory (from claim 16), a server system (from claims 1 and 12) and these are not sufficient to amount significantly more than the judicial exception—meaning, that there are no additional element(s) claimed in the dependent claims that could be significantly more than the judicial exception, but rather, further recites the abstract idea.
As indicated in the Step 2A Prong 2 analysis, the additional element(s) in the claims are merely, using a generic computer device or computing technologies and/or other machinery merely as a tool to perform an abstract idea that does not constitute a practical application and only amounts to a mere instruction to practice the invention. Thus, this does not render the claims as being eligible (refer to MPEP 2106.05(f) and 2106.05(h)). This is because the claimed invention must improve upon conventional functioning of a computer, or upon conventional technology or technological processes a technical explanation as to how to implement the invention should be present in the specification. The rationale set forth for the 2nd prong of the eligibility test above is also applicable and re-evaluated in the Step 2B analysis. Therefore, this rationale is sufficient for its rejection basis as it is not patent eligible and no comments are necessary as it is also consistent with MPEP 2106.
For dependent claims 2 – 11, 13 – 15 and 17 – 20, these claims cover or fall under the same abstract idea of a method of organizing human activity. They describe additional limitation steps of:
Claims 2 – 6, 13 – 15 and 17 – 19: describe training models using historical IVR transcripts and extracting subclassifications—i.e., routine machine learning (ML) techniques such as data labeling, sub-clustering, and creating embeddings
Claim 7: lists general categories of intent like billing or marketing, which are conceptual field of assistance, not technical elements
Claims 8 and 20: describe associating or routing the customer based on the classification—a mental step or automation of convention call center workflows
Claims 9 and 10: introduce communication with a chatbot using NLP transcription, which are well-known and generic functions
These dependent claims merely add further instructions to apply the abstract idea using general-purpose computer components and do not provide any inventive concept sufficient to overcome the abstract nature of the claimed subject matter. Thus, being directed to the abstract idea group of “managing personal behavior or relationships or interactions between people” and “commercial or legal interactions” as it is further handling the user data to route the user’s communication to the most appropriate/relevant team.
Step 2A Prong 2 and Step 2B: For dependent claims, these claims do not recite additional elements. However, the claim limitations are further describing the abstract idea and recite functions that amounts no more than mere instructions to apply the exception using a generic computer component and/or computing technologies (refer to MPEP 2106.05(f) and does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
Additionally, these elements and their limitations are “merely indicating a field of use or technological environment in which to apply a judicial exception do not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application” (MPEP 2106.05(h)). Therefore, the additional elements previously mentioned above, are nothing more than descriptive language about the elements that define the abstract idea, and these claims remain rejected under 101 as well.
Claim Rejections - 35 USC § 103
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.
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.
Claims 1 – 5, 7 – 8 and 16 - 20 are rejected under 35 U.S.C. 103 as being unpatentable over Wohlwend (US20200151254 A1) in view of Katz (US10135977 B1).
Regarding claim 1:
Wohlwend discloses:
receiving during a communication session, at a server system associated with the telecommunications network service provider, a communication from a customer, and Fig. 3; [¶0021 - 0022]: communication means by way of text messages or speech between a customer and a company.
wherein the communication session is a phone call between the customer and the telecommunications network service provider; and [¶0021] a customer may communicate by entering text messages or speaking, and the customer may send a message using any appropriate device, such as a computer, smart phone, tablet, wearable device, or Internet of things device.
generating, by the server system, a first embedding vector representing a numerical representation of natural language extracted from the communication; [¶0027]: Assigns an intent a ‘slot’ (i.e., a parameter or variable) in order to further determine the value of the slot of the message. Fig. 3; [¶0026]: Labels intents to categorize them by topics. Fig. 9; [¶0090]: Initial states of a message vector may be associated with a subset of possible intents.
comparing, by the server system, the first embedding vector to multiple embedding vectors stored in a vector database associated with the server system, Fig. 3; [¶0037]: the word embeddings of words may be constructed so that words with similar meanings or categories are close to one another in the N-dimensional vector space. The word embeddings for “cat” and “cats” may be close to each other because they have similar meanings, and the words “cat” and “dog” may be close to each other because they both relate to pets.
wherein each of the multiple embedding vectors is associated with an intent classification of multiple intent classifications, Fig. 3; [¶0043 – 0044]: Each stored vector is linked to a category referred to as an intent. The intents are then sorted and organized within the vector space.
wherein the multiple intent classifications correspond to a prediction of a type of assistance available for customers, Fig. 3; [¶0043]: The prototype comparison system determines the potential intents of the message. Additionally [¶0026]: labels the intents to organize and manage the intents to categories, i.e., “’PAY_BILL’ may include messages that express a desire to pay a bill”
wherein the multiple embedding vectors are created based on a model that is configured to identify customer communication intents, and Figs. 3, 8A – 8B; [¶0061]: The intent classifier of FIG. 3 may be applied to automating communications with a user. In some implementations, automated communications may use a graph (such as a directed graph or a tree) in guiding or structuring the receipt of information from the user. FIG. 8A illustrates an example graph that may be used for automated communications, and FIG. 8B illustrates example outgoing messages and intents for a communications system for handling airline reservations.
identifying, based on the comparison, which intent classification of the multiple intent classifications is associated with the first embedding vector, wherein the identification is performed in real-time during the communication session; and [¶0043]: Prototype comparison component 330 may receive the message embedding from message embedding component 320 and determine the intent of the message by comparing the message embedding with prototype vectors for possible intents.
redirecting, by configuring one or more communication links associated with the telecommunications network service provider, the phone call to a sub-unit of the telecommunications network service provider based on the identified intent classification; Figs. 5, 8B; [¶0062 – 0064]: Where a message matches an associated intent, processing may proceed to a further state of the graph. For example, the graph may have a root state that is used when starting a conversation with a user. In this example, the root state is marked as S1, and at state S1, the message “How can I help you today?” is presented to the user. Because the example application is for airline reservations, it may be expected that the user will respond with a message relating to one of three intents: (I1) BUY_PLANE_TICKET... Based on the response received from the user, processing may proceed to a next state in the graph. For example, for intent I1, processing may proceed to state S2, for intent 12, processing may proceed to state S3, and for intent 13, processing may proceed to state S4.
Wohlwend further discloses:
[¶0062]: A communication system trained to organize and identify the particular type of information obtained from a plethora of users. [¶0141]: The communication system being utilized by third-party companies over different network communications. [¶0062 – 0064]: Different intents of a user’s message corresponding to different states, with each state corresponding to a specific function (e.g., BUY, CANCEL, CHANGE). However, Wohlwend does not explicitly disclose the model being trained based on a set of interactive voice response (IVR) transcripts of historical customer communications, the communication link being associated with one of the communication link types, nor the sub-unit corresponding to network operations associated with the first communication link. Thus, Katz teaches:
wherein the model is trained based on a set of interactive voice response (IVR) transcripts of historical customer communications; [col. 2; line 22]: some embodiments, each corresponding model is based on a deep multimodal sequence auto-encoder (DMSA) that was trained with previous IVR transaction logs from the IVR system.
wherein the communication session includes a first communication link that is associated with the telecommunications network service provider and includes at least one of: a first frequency division duplex (FDD) operation, a first Time division duplex (TDD) operation, a first Long-Term Evolution (LTE) link, or a first millimeter wave link; [col. 7; line 65]: Data transmission of the IVR system may occur over a communication network. Alternatively, [col. 8; line 67]: The system may be connected through a plethora of communication networks. [Examiner’s Note]: Such telecommunication systems inherently operate over typical communication links, including cellular and wireless communication technologies. The recited communication link types (e.g., FDD, TDD, LTE, or millimeter wave) are well-known and generic implementations of telecommunication links used to support communication sessions. The claim does not require any specific configuration or use of these communication link types beyond their inclusion in the communication session. Therefore, under the broadest reasonable interpretation, Katz’s disclosure of receiving communications in an IVR telecommunications system inherently satisfies this limitation.
wherein the one or more configured communication links include at least one of: a second frequency division duplex (FDD) operation, a second Time division duplex (TDD) operation, a second Long-Term Evolution (LTE) link, or a second millimeter wave link, and; [col. 7; line 65]: Data transmission of the IVR system may occur over a communication network. Alternatively, [col. 8; line 67]: The system may be connected through a plethora of communication networks.
wherein the sub-unit of the telecommunications network service provider corresponds to network operations for the telecommunications network service provider associated with first communication link of the communication session. [col. 7; line 65]: Data transmission of the IVR system may occur over a communication network. Alternatively, [col. 4; line 60]: The IVR system comprising of an IVR menu, allowing for customer service operations (e.g., billing, account handling, service requests)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wohlwend’s disclosed prototype classifier communication system with that of a model trained based on a set of interactive voice response (IVR) transcripts of historical customer communications, telecommunication link types, as well as the sub-units corresponding to network operations associated with the first communication link, as taught by Katz, as using real-word/domain-specific data to improve model performance is a common and predictable technique in the field of machine learning and natural language processing. The motivation would be to improve the accuracy of the intent classification for customer service applications.
Regarding claim 2:
Wohlwend teaches all the limitations of claim 1, respectively, and further discloses:
associating, by the server system, the customer with the sub-unit of the telecommunications network service provider based on the identified intent classification, Fig. 3; [¶0061]: automated communications may use a graph (such as a directed graph or a tree) in guiding or structuring the receipt of information from the user. FIG. 8A illustrates an example graph that may be used for automated communications, and FIG. 8B illustrates example outgoing messages and intents for a communications system for handling airline reservations.
wherein the sub-unit of the telecommunications network service provider is associated with network operation, marketing, billing, or customer account management. Fig. 9; [¶0100]: customer of a company may contact the company to obtain assistance, such as to make a purchase, sign up for a subscription, change an address, request a copy of a bill, pay a bill, or request the status of a previous transaction.
Regarding claims 3 and 17:
Wohlwend teaches all the limitations of claims 1 - 2 and 20, respectively, and further discloses:
creating a first training set by: [¶0053]: The initial model m may be used to create an initial set of prototype vectors from the training data.
Furthermore, Wohlwend discloses:
[¶0024 – 0026]: The mathematical model may provide an output that indicates the intent of the message from a list of possible intents or that indicates that the message does not match any intent of the list of intents… an intent may be defined by a mathematical model that processes messages to determine intents of the messages or by a corpus of training data that was used to create the mathematical model. An intent may be assigned a label to make it easier for humans to understand the types of messages corresponding to the intent.) However, Wohlwend does not explicitly disclose the IVR element of the claim limitations. Thus, Katz teaches:
inputting the set of the IVR transcripts into a first model; and Fig. 3; [¶0068]: The model can be created by using the IVR transaction log as described above as inputs to the DMSA trained model
causing the first model to associate each of the IVR transcripts in the set of IVR transcripts with a respective intent classification of the multiple intent classifications to create the first training set, [¶0011]: the invention involves a method for optimization of interactive voice recognition (IVR) system processes.
wherein the first training set includes the set of the IVR transcripts, each of the IVR transcripts associated with the respective intent classification. [¶0011]: the invention involves a method for optimization of interactive voice recognition (IVR) system processes.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wohlwend’s disclosed prototype classifier communication system with that of the interactive voice recognition system, as using real-word/domain-specific data to improve model performance is a common and predictable technique in the field of machine learning and natural language processing. The motivation would be to improve the accuracy of the intent classification for customer service applications.
Regarding claims 4 and 18:
Wohlwend teaches all the limitations of claims 1 – 3, 17 and 20 respectively, and further discloses:
creating a second training set by: [¶0053]: The initial model m may be used to create an initial set of prototype vectors from the training data.
Furthermore, Wohlwend discloses:
[¶0024 – 0026]: “The mathematical model may provide an output that indicates the intent of the message from a list of possible intents or that indicates that the message does not match any intent of the list of intents… an intent may be defined by a mathematical model that processes messages to determine intents of the messages or by a corpus of training data that was used to create the mathematical model. An intent may be assigned a label to make it easier for humans to understand the types of messages corresponding to the intent.” Fig. 9; [¶0090]: The initial state may also be associated with one or more groups of intents where each group of intents is a subset of the possible intents from step 920. For example, a first group of intents may correspond to expected intents for messages received from the user at that state.) However, Wohlwend does not explicitly disclose the IVR element of the claim limitations. Thus, Katz teaches:
inputting the first training set to a second model; Fig. 3; [¶0068]: The model can be created by using the IVR transaction log as described above as inputs to the DMSA trained model” Additionally [¶0016]: each corresponding model is based on a deep multimodal sequence auto-encoder (DMSA) that was trained with previous IVR transaction logs from the IVR system.
extracting, by the second model, one or more sub-classifications for each of the IVR transcripts in the first training set; and Fig. 3; [¶0068]: The model can be created by using an IVR transaction log of IVR transactions captured by the same IVR system at a different period of time. Training the model can involve taking each vector created in step 350 above, and averaging all the values for all of the vectors, concatenating all of the vectors and inputting the average and the concatenated vector into the DMSA training module to create a trained model.
causing the second model to associate each of the IVR transcripts in the first training set with the one or more sub-classifications to create the second training set; and Fig. 3; [¶0068]: The model can be created by using an IVR transaction log of IVR transactions captured by the same IVR system at a different period of time. Training the model can involve taking each vector created in step 350 above, and averaging all the values for all of the vectors, concatenating all of the vectors and inputting the average and the concatenated vector into the DMSA training module to create a trained model.
creating, by a third model, the multiple embedding vectors from the second training set. Fig. 3; [¶0068]: The model can be created by using the IVR transaction log as described above as inputs to the DMSA trained model” Additionally [¶0016]: each corresponding model is based on a deep multimodal sequence auto-encoder (DMSA) that was trained with previous IVR transaction logs from the IVR system.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wohlwend’s disclosed system for identifying intents from user messages using a mathematical model trained with a corpus of training data and generating labeled classifications, where each message is assigned with an intent label to further assist and facilitate downstream processing, with a deep multimodal sequence auto encoder (DMSA) model using historical IVR system processes, as disclosed by Katz, in order to create a multi-stage machine learning system that classifies and processes customer communications derived from historical IVR transcripts. Such combination reflects a well-known design pattern in machine learning systems where embedding models are trained on structured labeled data to support downstream tasks (i.e., clustering, recommendation, or automated routing). Furthermore, such a combination would have produced predictable results by using known techniques for transcript classification and deep embedding generation, and would have merely involved the use of known techniques (auto encoding, classification) to improve a similar system (intent-based IVR transcript classification). Thus, the combination would have been obvious to a person of ordinary skill in the art at the time of the invention.
Regarding claim 5:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 4, 17 – 18 and 20, respectively, and Wohlwend further discloses:
wherein the first embedding vector is created based on a fourth model, and [¶0052 – 0053]: This mathematical model processes a sequence of word embeddings to compute a message embedding vector. The initial model m may be used to create an initial set of prototype vectors from the training data.
inputting, to the fourth model, the multiple embedding vectors from the second training set, [¶0025]: an intent may be defined by a mathematical model that processes messages to determine intents of the messages or by a corpus of training data that was used to create the mathematical model.
wherein each of the multiple embedding vectors includes a numerical representation of natural language extracted from the IVR transcripts and associated intent classification and one or more associated sub-classifications; and [¶0027]: Assigns an intent a ‘slot’ (i.e., a parameter or variable) in order to further determine the value of the slot of the message. Fig. 3; [¶0026]: Labels intents to categorize them by topics. Fig. 9; [¶0090]: Initial states of a message vector may be associated with a subset of possible intents.
training the fourth model to create embedding vectors based on natural language input. Fig. 3; [¶0050 – 0051]: Mathematical models are trained using a corpus of training data in order to then be used by message embedding components.
Regarding claim 7:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 6, 17 – 18 and 20. Wohlwend further discloses:
wherein the intent classifications include an intent related to network operation, an intent related to marketing, an intent related to billing, or an intent related to customer accounts. Fig. 3; [¶0026]: Labels intents to categorize them by topics, including user’s needs.
Regarding claim 8:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 7, 17 – 18 and 20. Wohlwend further discloses:
wherein each intent classification includes two or more sub-classifications, [¶0084]: Intents can be further divided into two or more groups.
wherein each of the multiple embedding vectors is further associated with a sub-classification of the two or more sub-classifications, and [¶0084]: Message embeddings are compared with all vectors for a first group of intents to find a match. If no match is found, subsequent intent sub-groups are then matched.
wherein the method further includes identifying which intent sub-classification of the multiple intent sub-classifications is associated with the first embedding vector. Figs. 3 – 5 and 8B – 9; [¶0094 – 0095]: Intents are selected and then compared to find one or more intents in order to 1) find expected intents of the user 2) correspond to the user’s current request 3) correspond to intents that don’t relate to the user’s request. [Examiner’s Note: In view of BRI, the claim is interpreted as determining the general category of the user’s request as well as the specific reason for the call/request.]
Regarding claim 10:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 8, 17 – 18 and 20, respectively, and Wohlwend further discloses:
in an instance that the communication includes oral communication, creating, by natural language processing (NLP), a transcript of the communication comprising the natural language. Figs. 4 – 5; [¶0037 - 0038]: Word embeddings are computed using natural language stored in a training corpus through different techniques and software (i.e., Word2Vec, GloVe) and then stored in a vector space.
Regarding claim 11:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 10, 13 – 15, 17 – 18 and 20., respectively, and Wohlwend further discloses:
wherein creating the first embedding vector includes parsing the natural language of the communication into a sequence of text segments, and Figs. 3 and 5; [¶0037 – 0041]: The word embedding component processes the natural language message into word embeddings for each of the words within the message.
converting the sequence of text segments into the numerical representation of the natural language used for creating the first embedding vector. [¶0027]: Assigns an intent a ‘slot’ (i.e., a parameter or variable) in order to further determine the value of the slot of the message. Fig. 3; [¶0026]: Labels intents to categorize them by topics. Fig. 9; [¶0090]: Initial states of a message vector may be associated with a subset of possible intents.
Regarding claim 16:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 11, 17 – 18 and 20, respectively, and Wohlwend further discloses:
at least one hardware processor; and Fig. 14; [¶0144]: teaches “Computing device 1400 may include any components typical of a computing device, such as, one or more processors 1411, and one or more network interfaces 1412.
at least one non-transitory memory storing instructions, which, when executed by the at least one hardware processor, cause the system to: Fig. 14; [¶0146]: teaches “Computing device 1400 may include or have access to various data stores. Data stores may use any known storage technology such as files, relational databases, non-relational databases, or any non-transitory computer-readable media.
receive, during a communication session, a communication from a customer associated with the telecommunications network, [¶0021]: a customer may communicate by entering text messages or speaking, and the customer may send a message using any appropriate device, such as a computer, smart phone, tablet, wearable device, or Internet of things device.
wherein the communication includes natural language, and [¶0020]: teaches “A user may interact with computers or automated services using natural language.
wherein the communication session is between the customer and a customer service representative associated with the telecommunications network service provider; [¶0021]: A customer may seek support from a company using a variety of communication techniques, and the techniques described herein are not limited to any particular communication techniques. For example, a customer may communicate by entering text messages or speaking, and the customer may send a message using any appropriate device, such as a computer, smart phone, tablet, wearable device, or Internet of things device.
generate a first embedding vector based on the natural language of the communication, [¶0027]: A natural language message is received, to which it is then processed through a mathematical model to determine the intent of the message. The system then assigns an intent a ‘slot’ (i.e., a parameter or variable) in order to further determine the value of the slot of the message. Fig. 3; [¶0035 - 0036]: Prototype vectors are utilized in order to determine an intent of the natural language message.
wherein the first embedding vector includes a numerical representation of natural language extracted from the communication; [¶0027]: Assigns an intent a ‘slot’ (i.e., a parameter or variable) in order to further determine the value of the slot of the message. Fig. 3; [¶0026]: Labels intents to categorize them by topics. Fig. 9; [¶0090]: Initial states of a message vector may be associated with a subset of possible intents.
compare the first embedding vector to multiple embedding vectors stored in a vector database associated with the system, Figs. 3 - 7 [¶0049]: Message embeddings are compared to prototype vectors using hierarchical techniques. Multiple levels of clustering may also be used in the process of comparing a message embedding with a cluster. Fig. 14; [¶0146]: A computing device holds a prototype vectors data store which is then configured to a communication system.
wherein each of the multiple embedding vectors is associated with an intent classification of multiple intent classifications correspond to a prediction of a type of assistance available for customers, and Fig. 3; [¶0043 – 0044]: Each stored vector is linked to a category referred to as an intent. The intents are then sorted and organized within the vector space.
wherein the multiple embedding vectors are created based on a model that is configured to identify customer communication intent; Figs. 3, 8A – 8B; [¶0061]: The intent classifier of FIG. 3 may be applied to automating communications with a user. In some implementations, automated communications may use a graph (such as a directed graph or a tree) in guiding or structuring the receipt of information from the user. FIG. 8A illustrates an example graph that may be used for automated communications, and FIG. 8B illustrates example outgoing messages and intents for a communications system for handling airline reservations.
identify based on the comparison, which intent classification of the multiple intent classifications is associated with the first embedding vector; and [¶0043]: Prototype comparison component 330 may receive the message embedding from message embedding component 320 and determine the intent of the message by comparing the message embedding with prototype vectors for possible intents.
redirect, by configuring one or more communication links associated with the telecommunications network service provider, the communication session to a sub- unit of the telecommunications network service provider based on the identified intent classification Figs. 5, 8B; [¶0062 – 0064]: Where a message matches an associated intent, processing may proceed to a further state of the graph. For example, the graph may have a root state that is used when starting a conversation with a user. In this example, the root state is marked as S1, and at state S1, the message “How can I help you today?” is presented to the user. Because the example application is for airline reservations, it may be expected that the user will respond with a message relating to one of three intents: (I1) BUY_PLANE_TICKET... Based on the response received from the user, processing may proceed to a next state in the graph. For example, for intent I1, processing may proceed to state S2, for intent 12, processing may proceed to state S3, and for intent 13, processing may proceed to state S4.
Wohlwend further discloses:
[¶0062]: A communication system trained to organize and identify the particular type of information obtained from a plethora of users. [¶0141]: The communication system being utilized by third-party companies over different network communications. [¶0062 – 0064]: Different intents of a user’s message corresponding to different states, with each state corresponding to a specific function (e.g., BUY, CANCEL, CHANGE). However, Wohlwend does not explicitly disclose the model being trained based on a set of interactive voice response (IVR) transcripts of historical customer communications, the communication link being associated with one of the communication link types, nor the sub-unit corresponding to network operations associated with the first communication link. Thus, Katz teaches:
wherein the communication session includes a first communication link that is associated with the telecommunications network service provider and includes at least one of: a first FDD operation, a first TDD operation, a first LTE link, or a first millimeter wave link; [col. 7; line 65]: Data transmission of the IVR system may occur over a communication network. Alternatively, [col. 8; line 67]: The system may be connected through a plethora of communication networks. [Examiner’s Note]: As claim 1.
wherein the one or more configured communication links include at least one of: a second FDD operation, a second LTE link, or a second millimeter wave link, and [col. 7; line 65]: Data transmission of the IVR system may occur over a communication network. Alternatively, [col. 8; line 67]: The system may be connected through a plethora of communication networks.
wherein the sub-unit of the telecommunications network service provider corresponds to network operations for the telecommunications network service provider associated with the first communication link of the communication session. [col. 7; line 65]: Data transmission of the IVR system may occur over a communication network. Alternatively, [col. 4; line 60]: The IVR system comprising of an IVR menu, allowing for customer service operations (e.g., billing, account handling, service requests)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wohlwend’s disclosed prototype classifier communication system with that of a model trained based on a set of interactive voice response (IVR) transcripts of historical customer communications, telecommunication link types, as well as the sub-units corresponding to network operations associated with the first communication link, as taught by Katz, as using real-word/domain-specific data to improve model performance is a common and predictable technique in the field of machine learning and natural language processing. The motivation would be to improve the accuracy of the intent classification for customer service applications.
Regarding claim 19:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 18 and 20, respectively, and Wohlwend further discloses:
wherein the first embedding vector is created based on a fourth model, and [¶0052 – 0053]: This mathematical model processes a sequence of word embeddings to compute a message embedding vector. The initial model m may be used to create an initial set of prototype vectors from the training data.
the system is further caused to train the fourth model by: Fig. 3; [¶0050 – 0051]: Mathematical models are trained using a corpus of training data in order to then be used by message embedding components.
inputting, to the fourth model, the multiple embedding vectors from the second training set, [¶0025]: an intent may be defined by a mathematical model that processes messages to determine intents of the messages or by a corpus of training data that was used to create the mathematical model.
wherein each of the multiple embedding vectors includes a numerical representation of natural language extracted from the IVR transcripts and associated intent classification and one or more associated sub-classifications; and [¶0027]: Assigns an intent a ‘slot’ (i.e., a parameter or variable) in order to further determine the value of the slot of the message. Fig. 3; [¶0026]: Labels intents to categorize them by topics. Fig. 9; [¶0090]: Initial states of a message vector may be associated with a subset of possible intents.
training the fourth model to create embedding vectors based on natural language input. Fig. 3; [¶0050 – 0051]: Mathematical models are trained using a corpus of training data in order to then be used by message embedding components.
Regarding claim 20:
Wohlwend discloses:
associating, by the system, the customer with the sub-unit of the telecommunications network service provider based on the identified intent classification, Fig. 3; [¶0061]: automated communications may use a graph (such as a directed graph or a tree) in guiding or structuring the receipt of information from the user. FIG. 8A illustrates an example graph that may be used for automated communications, and FIG. 8B illustrates example outgoing messages and intents for a communications system for handling airline reservations.
wherein the sub-unit of the telecommunications network service provider is associated with network operation, marketing, billing, or customer account management. Fig. 9; [¶0100]: customer of a company may contact the company to obtain assistance, such as to make a purchase, sign up for a subscription, change an address, request a copy of a bill, pay a bill, or request the status of a previous transaction.
Claims 6, 9 and 12 - 15 are rejected under 35 U.S.C. 103 as being unpatentable over Wohlwend (US20200151254 A1) in view of Katz (US10135977 B1) in further view of Schuetz (US20230412530 A1).
Regarding claim 6:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 5, 13 – 15, 17 – 18 and 20. However, neither Wohlwend nor Katz disclose either of the cosine distancing limitations. Thus, Schuetz teaches:
determining cosine distances between the first embedding vector and the multiple embedding vectors, Fig. 1 [¶0034]: A system calculates the cosine distance by transforming natural language expressions into vectors using sentence embedding.
wherein the cosine distances represent similarities between the first embedding vector and the multiple embedding vectors. Fig. 1; [¶0034 – 0035]: In order to calculate cosine distance, the system computes cosine similarities between vectors of the first and second chatbots as well as the angles between them.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Wohlwend’s disclosed method and system for processing communications using a prototype classifier with determining cosine distances between the first embedding vector and the multiple embedding vectors as well as the cosine distances representing similarities between the first embedding vectors and the multiple embedding vectors, as taught by Schuetz, as “training data sets of the individual chatbots do not fit together very well for a digital assistant scenario” (Abstract, Schuetz).
Regarding claim 9:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 8, 17 – 18 and 20. However, neither Wohlwend nor Katz disclose either of the cosine distancing limitations. Thus, Schuetz teaches:
wherein the communication is received by a customer service representative or a chatbot. Fig. 1; [¶0010]: A chatbot is selected from a plurality of chatbots to respond to a user’s input based on which is best-fitted to the request.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Wohlwend’s disclosed method and system for processing communications using a prototype classifier with wherein the communication is received by a customer service representative or a chatbot, as taught by Schuetz, as “training data sets of the individual chatbots do not fit together very well for a digital assistant scenario” (Abstract, Schuetz).
Regarding claim 12:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 11, 13 – 15, 17 – 18 and 20, respectively, and Wohlwend further discloses:
creating, by the server system, a first embedding vector based on the natural language of the communication, wherein the first embedding vector includes a numerical representation of natural language extracted from the communication; Fig. 3; [¶0037]: a message is received. Word embedding component 310 may process the message to obtain a word embedding for each word of the message. A word embedding is a vector in an N-dimensional vector space that represents the word but does so in a manner that preserves useful information about the meaning of the word.
comparing, by the server system, the first embedding vector to multiple embedding vectors stored in a vector database associated with the server system, Fig. 3; [¶0037]: word embeddings of words may be constructed so that words with similar meanings or categories are close to one another in the N-dimensional vector space. For example, the word embeddings for “cat” and “cats” may be close to each other because they have similar meanings, and the words “cat” and “dog” may be close to each other because they both relate to pets.
wherein each of the multiple embedding vectors is associated with an intent classification of multiple intent classifications, Fig. 3; [¶0043 – 0044]: Each stored vector is linked to a category referred to as an intent. The intents are then sorted and organized within the vector space.
wherein the multiple embedding vectors are created based on a model that is configured to identify customer communication intent, and Fig. 3, 8A – 8B; [¶0061]: The intent classifier of FIG. 3 may be applied to automating communications with a user. In some implementations, automated communications may use a graph (such as a directed graph or a tree) in guiding or structuring the receipt of information from the user. FIG. 8A illustrates an example graph that may be used for automated communications, and FIG. 8B illustrates example outgoing messages and intents for a communications system for handling airline reservations.
Wohlwend discloses a prototype comparison component having the capabilities to analyze a user’s message and determine which state/node the message belongs to in regards to its intent [¶0043]. Katz teaches a communications link of network providers including a variety of network link types [col. 7; line 65] and [col. 8; line 67]. Additionally, Katz also teaches redirecting communication methods to a sub-unit of different service providers based on an identified intent classification [Figs. 5, 8B; ¶0062 – 0064]. However, neither Wohlwend nor Katz teach the chatbot elements of the claim limitations, however, Schuetz teaches:
receiving, by a server system associated with the telecommunications network service provider during a chatbot session, a communication from a customer associated with the telecommunications network, Fig. 1; [¶0010]: A chatbot is selected from a plurality of chatbots to respond to a user’s input based on which is best-fitted to the request.
wherein the communication includes natural language, and [¶0021 - 0022]: communication means by way of text messages or speech between a customer and a company.
wherein the chatbot session is between the customer and a chatbot software application configured to generate natural language in response to communications received from customers; Figs. 2 – 3; [¶0016]: A bot orchestrator selects a chatbot from a plethora of chatbots to respond with the best fit answer to the user’s request in natural language.
identifying, based on the comparison, which intent classification of the multiple intent classifications is associated with the first embedding vector, wherein the identification is performed in real-time during the chatbot session; and Fig. 1; [¶0010]: A chatbot is selected from a plurality of chatbots to respond to a user’s input based on which is best-fitted to the request.
generating, by the chatbot software application, a response to the received communication based on the identified intent classification. Figs. 2 – 3; [¶0020]: A user’s message triggers the chatbot orchestrator to filter and categorize the intent of the message to then respond with an appropriate expression.
wherein the chatbot session is associated with a first communication link of a telecommunications network service provider and includes at least one of a first frequency division duplex (FDD) operation, a first Time division duplex (TDD) operation, a first Long-Term Evolution (LTE) link, or a first millimeter wave link; Fig. 1; [¶0010]: A chatbot is selected from a plurality of chatbots to respond to a user’s input based on which is best-fitted to the request.
redirecting, by configuring one or more communication links associated with the telecommunications network service provider, the chatbot session to a sub-unit of the telecommunications network service provider based on the identified intent classification, Fig. 1; [¶0010]: A chatbot is selected from a plurality of chatbots to respond to a user’s input based on which is best-fitted to the request.
wherein the one or more configured communication links include at least one of: a second frequency division duplex (FDD) operation, a second Time division duplex (TDD) operation, a second Long-Term Evolution (LTE) link, or a second millimeter wave link, and Fig. 1; [¶0010]: A chatbot is selected from a plurality of chatbots to respond to a user’s input based on which is best-fitted to the request.
wherein the sub-unit of the telecommunications network service provider corresponds to network operations for the telecommunications network service provider associated with the first communication link of the chatbot session. Fig. 1; [¶0010]: A chatbot is selected from a plurality of chatbots to respond to a user’s input based on which is best-fitted to the request.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Wohlwend’s disclosed method and system for processing communications using a prototype classifier with communication link types, as taught by Katz, with the systems and methods of a chatbot, as taught by Schuetz, in order to increase efficiency and convenience during communication between a customer and company in regards to a customer’s request. Such modification would have been a predictable use of a known technique to improve a similar system in the same way.
Regarding claim 13:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 12, 16 – 18 and 20. However, neither Wohlwend nor Katz disclose either of the cosine distancing limitations. Thus, Schuetz teaches:
creating a first training set by: [¶0053]: The initial model m may be used to create an initial set of prototype vectors from the training data.
Furthermore, Wohlwend discloses:
[¶0024 – 0026]: The mathematical model may provide an output that indicates the intent of the message from a list of possible intents or that indicates that the message does not match any intent of the list of intents… an intent may be defined by a mathematical model that processes messages to determine intents of the messages or by a corpus of training data that was used to create the mathematical model. An intent may be assigned a label to make it easier for humans to understand the types of messages corresponding to the intent.) However, Wohlwend does not explicitly disclose the IVR element of the claim limitations. Thus, Katz teaches:
inputting the set of the IVR transcripts into a first model; and Fig. 3; [¶0068]: The model can be created by using the IVR transaction log as described above as inputs to the DMSA trained model
causing the first model to associate each of the IVR transcripts in the set of IVR transcripts with a respective intent classification of the multiple intent classifications to create the first training set, [¶0011]: the invention involves a method for optimization of interactive voice recognition (IVR) system processes.
wherein the first training set includes the set of the IVR transcripts, each of the IVR transcripts associated with the respective intent classification. [¶0011]: the invention involves a method for optimization of interactive voice recognition (IVR) system processes.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wohlwend’s disclosed prototype classifier communication system with that of the interactive voice recognition system, as using real-word/domain-specific data to improve model performance is a common and predictable technique in the field of machine learning and natural language processing. The motivation would be to improve the accuracy of the intent classification for customer service applications.
Regarding claim 14:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 5, 13 – 15, 17 – 18 and 20. However, neither Wohlwend nor Katz disclose either of the cosine distancing limitations. Thus, Schuetz teaches:
creating a second training set by: [¶0053]: The initial model m may be used to create an initial set of prototype vectors from the training data.
Furthermore, Wohlwend discloses:
[¶0024 – 0026]: “The mathematical model may provide an output that indicates the intent of the message from a list of possible intents or that indicates that the message does not match any intent of the list of intents… an intent may be defined by a mathematical model that processes messages to determine intents of the messages or by a corpus of training data that was used to create the mathematical model. An intent may be assigned a label to make it easier for humans to understand the types of messages corresponding to the intent.” Fig. 9; [¶0090]: The initial state may also be associated with one or more groups of intents where each group of intents is a subset of the possible intents from step 920. For example, a first group of intents may correspond to expected intents for messages received from the user at that state.) However, Wohlwend does not explicitly disclose the IVR element of the claim limitations. Thus, Katz teaches:
inputting the first training set to a second model; Fig. 3; [¶0068]: The model can be created by using the IVR transaction log as described above as inputs to the DMSA trained model” Additionally [¶0016]: each corresponding model is based on a deep multimodal sequence auto-encoder (DMSA) that was trained with previous IVR transaction logs from the IVR system.
extracting, by the second model, one or more sub-classifications for each of the IVR transcripts in the first training set; and Fig. 3; [¶0068]: The model can be created by using an IVR transaction log of IVR transactions captured by the same IVR system at a different period of time. Training the model can involve taking each vector created in step 350 above, and averaging all the values for all of the vectors, concatenating all of the vectors and inputting the average and the concatenated vector into the DMSA training module to create a trained model.
causing the second model to associate each of the IVR transcripts in the first training set with the one or more sub-classifications to create the second training set; and Fig. 3; [¶0068]: The model can be created by using an IVR transaction log of IVR transactions captured by the same IVR system at a different period of time. Training the model can involve taking each vector created in step 350 above, and averaging all the values for all of the vectors, concatenating all of the vectors and inputting the average and the concatenated vector into the DMSA training module to create a trained model.
creating, by a third model, the multiple embedding vectors from the second training set. Fig. 3; [¶0068]: The model can be created by using the IVR transaction log as described above as inputs to the DMSA trained model” Additionally [¶0016]: each corresponding model is based on a deep multimodal sequence auto-encoder (DMSA) that was trained with previous IVR transaction logs from the IVR system.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wohlwend’s disclosed system for identifying intents from user messages using a mathematical model trained with a corpus of training data and generating labeled classifications, where each message is assigned with an intent label to further assist and facilitate downstream processing, with a deep multimodal sequence auto encoder (DMSA) model using historical IVR system processes, as disclosed by Katz, in order to create a multi-stage machine learning system that classifies and processes customer communications derived from historical IVR transcripts. Such combination reflects a well-known design pattern in machine learning systems where embedding models are trained on structured labeled data to support downstream tasks (i.e., clustering, recommendation, or automated routing). Furthermore, such a combination would have produced predictable results by using known techniques for transcript classification and deep embedding generation, and would have merely involved the use of known techniques (auto encoding, classification) to improve a similar system (intent-based IVR transcript classification). Thus, the combination would have been obvious to a person of ordinary skill in the art at the time of the invention.
Regarding claim 15:
The combination of Wohlwend and Katz teaches all the limitations of claims 1 – 5, 13 – 15, 17 – 18 and 20. However, neither Wohlwend nor Katz disclose either of the cosine distancing limitations. Thus, Schuetz teaches:
wherein the first embedding vector is created based on a fourth model, and [¶0052 – 0053]: This mathematical model processes a sequence of word embeddings to compute a message embedding vector. The initial model m may be used to create an initial set of prototype vectors from the training data.
inputting, to the fourth model, the multiple embedding vectors from the second training set, [¶0025]: an intent may be defined by a mathematical model that processes messages to determine intents of the messages or by a corpus of training data that was used to create the mathematical model.
wherein each of the multiple embedding vectors includes a numerical representation of natural language extracted from the IVR transcripts and associated intent classification and one or more associated sub-classifications; and [¶0027]: Assigns an intent a ‘slot’ (i.e., a parameter or variable) in order to further determine the value of the slot of the message. Fig. 3; [¶0026]: Labels intents to categorize them by topics. Fig. 9; [¶0090]: Initial states of a message vector may be associated with a subset of possible intents.
training the fourth model to create embedding vectors based on natural language input. Fig. 3; [¶0050 – 0051]: Mathematical models are trained using a corpus of training data in order to then be used by message embedding components.
Pertinent Art
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure.
Mehrotra (US20240143945 A1) is pertinent because it is directly related to “natural language processing and machine learning systems, and more specifically to systems and methods for intent classification to provide natural language understanding in a conversation agent.”
Solis (US12265576 B2) is pertinent because it is directed to “natural language processing, and more particularly, to systems and methods for multilingual intent classification predictions.”
Brdiczka (US20210303784 A1) is pertinent because it is related to “systems, non-transitory computer-readable media, and methods for improving accuracy of natural language input classification by using heuristics and machine learning to determine when to process instances of natural language input.”
Jonnalagadda (US20200143247 A1) is pertinent because it is directly related to “systems and methods for natural language processing and generation of more “human” sounding artificially generated conversations. Such natural language processing techniques may be employed in the context of machine learned conversation systems.”
Semeniuk (US11777874 B1) is pertinent because it is directed to “systems and methods for automated conversations. More particularly, the present disclosure relates to an artificial intelligence-based conversation engine. Even more specifically, the present disclosure relates to an artificial intelligence-based conversation engine trained to use intent to provide a natural conversation experience.”
Venkataraman (US12183344 B1) is pertinent because it is related to “predicting an entity and intent based on captured speech and, more particularly, to systems and methods for determining a next action based on one or more predicted entities and one or more predicted intents, the one or more predicted entities and the one or more predicted intents based on captured speech.”
Darla (US20240386883 A1) is pertinent because it is directly related to “systems and methods for intent prediction and usage.”
Ilagan (US20120320905 A1) is pertinent because it is directed to “information handling systems, and more particularly relates to a system and a method for providing customer support using an information handling system.”
Kaniganti (US20240089372 A1) is pertinent because it is related to “A call center is a managed capability that can be centralized or remote and is used for receiving or transmitting a large volume of inquiries by telephone. An inbound call center is operated by an entity to administer incoming product or service support or information inquiries from consumers.”
Britt (US20240144088 A1) is pertinent because it is directly related to “tools enabled by machine learning to perform interaction analysis and summarization, agent training and chat analytics to provide automated coaching enhancement via robotic training, and more particularly, robot-enabled training delivery to a human customer service agent, automated coaching and meeting analytics, and an employee engagement, retention, churn, and prediction platform.”
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Bill Chen whose telephone number is (571)270-0660. The examiner can normally be reached Monday - Friday 8:30am - 5:00pm.
THIS ACTION IS MADE FINAL. 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.
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, Nathan Uber can be reached on (571) 270-3923. 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.
/BILL CHEN/Examiner, Art Unit 3626
/NATHAN C UBER/Supervisory Patent Examiner, Art Unit 3626
Prosecution Insights