DETAILED ACTION
This office action is in response to Applicant’s Amendment/Request for Reconsideration, received on 02/26/2026. Claims 1, 3, 10, 12, and 19 have been amended. Claims 1-4, 7-13, 16-23 are pending and have been considered.
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 .
Response to Arguments
Applicant's arguments filed 02/26/2026 have been fully considered but they are not persuasive.
Applicant’s representative asserts, “On page 11 of the Office Action, the Examiner cites Rahman to reject inter alia the limitations:
based on [a] retrieved at least one case, generating . . . a constraint that comprises an if-then condition ....
However, Applicant notes that a careful inspection of the cited prior art reveals that Rahman does not actually disclose these limitations. For example, in the Office Action, the Examiner rejects these limitations on the ground that paragraph [0046] of Rahman discloses the following:
‘if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule…’
However, Applicant disagrees with this ground of rejection because although this excerpt may arguably be considered to disclose ‘a constraint that comprises an if-then condition,’ Applicant respectfully submits that the cited excerpt does not provide any disclosure whatsoever of the actual ‘generation’ of Rahman's alleged ‘constraint,’ far less any disclosure that the generation is ‘based on ... at least one case ....’
Therefore, Applicant respectfully submits that Rahman cannot be reasonably considered to disclose the above-mentioned rejected limitations because as stated above, the cited disclosure is noticeably silent on the limitations' ‘generating’ feature, which is explicitly required by the language of these limitations. Indeed, the very title of the present application is explicitly drawn to inter alia the ‘generation of rules ....’”
In response, the examiner would like to refer to the broadest reasonable interpretation of the claim language “generating…a constraint that comprises and if-then condition” in view of the cited art. Specifically, the cited portion of Rahman discloses “if-then” condition based on similarity of documents to template structures as agreed with by Applicant. With regard to the generation, the examiner would like to refer to Fig. 4 of Rahman, “Rule Generator 408”. [0055] of Rahman discloses “historical-based rules 540 may be generated and/or analyzed via rule generator 408 of FIG. 4. For instance, historical-based rules 540 may be used as a basis for generating/deriving new rules to be added to a set of rules associated with a particular domain”. Taking this disclosure in view of the similarity score if-then, wherein the “template matching rule” indicates a generated rule based on how similar documents are as would be performed by the rule generator 408. The examiner respectfully asserts that Rahman does disclose generating constraints, i.e. rules. Further, if the constraint generated is a similarity constraint, this indicates that the retrieved case to be used for a similarity comparison to a current case is “based on” the retrieved case as is required for a similarity between the retrieved case and a current case.
Applicant’s representative continues, “On page 9 of the Office Action, the Examiner cites Sethi to reject inter alia the limitation a ‘machine learning model of a randomly changing system…’ However, Applicant notes that a careful inspection of the cited prior art reveals that Sethi does not actually disclose this rejected limitation. For example, the Examiner cites Sethi to reject this limitation on the following grounds: [0127] training samples may be based on one or more of random selection, [0147] FIG. 8 illustrates an example artificial neural network (‘ANN’) 800. In particular embodiments, an ANN may refer to a computational model ....
However, Applicant disagrees with these grounds of rejection because although Sethi discloses that its ‘training samples may be based on ... random selection,’ Applicant respectfully submits that training samples ... based on ..[a] random selection’ are not the same as ‘a randomly changing system,’ far less a ‘machine learning model of a randomly changing system ....’
Indeed, the mere generation of Sethi's ‘training samples’ cannot reasonably be considered to ‘model [] a randomly changing system’ or to be a ‘machine learning model’ itself. Therefore, Applicant respectfully submits that Sethi cannot be reasonably considered to disclose a ‘machine learning model of a randomly changing system ....’
Therefore, to cure this clear deficiency of Sethi, the Examiner argues the following:
Disclosing an ANN which refers to a model while also disclosing that the model may contain Markov models indicates the ANN to be comprising a Markov machine learning model of a randomly changing system, i.e. during at least a training.
However, Applicant notes that the Examiner presents these conclusory arguments without any objective support. Accordingly, Applicant respectfully submits that this ground of rejection is patently improper because it fails to provide any objective evidence whatsoever for the argument that:
disclosing that [a] model may contain Markov models indicates . . . a Markov machine learning model of a randomly changing system, i.e. during at least a training.
Therefore, Applicant respectfully submits that this conclusory argument cannot reasonably be considered to cure the above-mentioned deficiency of Sethi.”
In response, the examiner would like to refer to the broadest reasonable interpretation of the claim language as currently presented in view of the cited art. Specifically, the examiner respectfully asserts that a “model that comprises a Markov machine learning model of a randomly changing system” does not indicate that the model itself has to be randomly changing. As inputs to the same model change, the system comprising the model and input data will necessarily be changing as the input data is randomized. Sethi explicitly discloses an ASR module which may include a combination of hidden Markov models and neural networks ([0106]). Indicating a combination of these elements within the same ASR module indicates the ASR module to be a model of a randomly changing system as the training input is randomly selected, necessarily affecting, i.e. randomizing, how processing is performed based on received input (consider both speech and non-speech as input). Further, Fig. 2, decision point D0 205 decides “whether to begin processing the user input in the first operational mode (i.e., on-device mode), the second operational mode (i.e., cloud mode), or the third operational mode (i.e., blended mode)” ([0046]). This indicates a randomly changing system (in terms of operational mode) based on a randomly received training input, wherein the system clearly utilizes at least a combination of ANNs and Markov models, wherein an ASR module containing both of these types of models suggests the ASR module/model to be generated based at least on utilized ANN techniques, comprising a Markov machine learning model.
Applicant's arguments fail to comply with 37 CFR 1.111(b) because they amount to a general allegation that the claims define a patentable invention without specifically pointing out how the language of the claims patentably distinguishes them from the references.
Applicant's arguments do not comply with 37 CFR 1.111(c) because they do not clearly point out the patentable novelty which he or she thinks the claims present in view of the state of the art disclosed by the references cited or the objections made. Further, they do not show how the amendments avoid such references or objections.
Applicant’s representative continues, “On pages 36-37 of the Office Action, the Examiner acknowledges that ‘Rahman in view of Sethi, further in view of Bonissone does not disclose’ inter alia the limitations:
wherein the at least one annotation includes ... a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept ....
However, on pages 37-38 of the Office Action, the Examiner relies on Mukherjee to reject these limitations. In response, Applicant notes that a careful inspection of the cited prior art reveals that Mukherjee does not actually disclose these rejected limitations.
For example, at the bottom of page 37 of the Office Action, the Examiner cites Mukherjee to reject the foregoing limitations on the following grounds:
Fig. 10, Document Array (D), Col. 9, Lines 56-59 - In various aspects, one or more of, or all of, the plurality of text blocks that were retrieved for the records sorted into the primary cluster are stored by the rule engine 104 in array ‘D’....
(Emphasis added.) However, Applicant respectfully submits that merely storing ‘text blocks that were retrieved’ cannot reasonably be considered to be the same as appending a ‘concept ... to a numeric suffix that accounts for each occurrence of the [] concept ... .’ Indeed, Mukherjee provides no disclosure whatsoever of ‘a numeric suffix that accounts for each occurrence of...[a] concept,’ far less of an appendage of a ‘corresponding concept’ to such ‘a numeric suffix.’
Therefore, to cure this clear deficiency of Mukherjee, the Examiner argues that:
The symptoms/historical medical events gathered from the patient input of Fig. 7 being represented in an array, combined with the term/classification labels, i.e. the annotation, indicates the combination to be an appending of a numerical suffix, i.e. the classification, with a corresponding concept, i.e. the symptom/past medical history event, wherein each symptom having an associated classification, i.e. numeric suffix, indicates the numeric suffixes to be accounting for each occurrence of the concept (consider the two instances of classification ‘R10.9’ in document array D). The examiner would like to note that a ‘numeric suffix that accounts for each occurrence of the corresponding concept’ does not necessarily require an interpretation of that numeric suffix to be an overall count of total instances of the associated concepts, though the examiner believes this to be Applicant's intent. Further, consider the ‘term frequency’ metric of Mukherjee which suggest an overall count of classifications, Col. 9, Lines 60-63.
However, Applicant notes that the Examiner presents these conclusory arguments without any objective support. Accordingly, Applicant respectfully submits that these grounds of rejection are patently improper because they fail to provide any objective evidence for the argument that:
symptoms ... represented in an array, combined with [a] term ... indicates ... an appending of a numerical suffix ....
Additionally, Applicant also submits that the Examiner's rejection does not provide any objective support for the argument that ‘an associated classification’ is the same as a ‘numeric suffix.’ Moreover, Applicant further submits that the Examiner's rejection fails to provide any objective support for the arguments that: (1) symptoms ‘having an associated classification’ indicates ‘accounting for each occurrence’; and (2) ‘the “term frequency” metric of Mukherjee [] suggest an overall count of classifications ....’
Therefore, Applicant respectfully submits that this conclusory argument cannot reasonably be considered to cure the above-mentioned deficiency of Mukherjee.”
In response, the examiner would like to refer to Mukherjee in view of the BRI of the claim language as currently presented. Specifically, the examiner respectfully disagrees with Applicant’s assertion that the examiner does not provide objective support for the argument that “an associated classification” is the same as a “numeric suffix”. The document array of Fig. 10 of Mukherjee containing symptoms with associated identifiers (containing numbers after the symptom is listed) indicates the identifier to be a numeric suffix which accounts for each symptom/medical history episode with as associated numeric suffix. The examiner asserts that the array D is containing annotations including a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept as currently claimed. It is unclear to the examiner how “an associated classification” does not track to a “numeric suffix” when the classification contains numbers and is after the text, i.e. a numeric suffix.
Applicant's arguments fail to comply with 37 CFR 1.111(b) because they amount to a general allegation that the claims define a patentable invention without specifically pointing out how the language of the claims patentably distinguishes them from the references.
Applicant's arguments do not comply with 37 CFR 1.111(c) because they do not clearly point out the patentable novelty which he or she thinks the claims present in view of the state of the art disclosed by the references cited or the objections made. Further, they do not show how the amendments avoid such references or objections.
The rejections of independent claims will be updated to incorporate the previous subject matter of dependent claim 3 (and associated equivalents) using Mukherjee. See updated rejections below.
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.
Claim(s) 1-4, 7-13, 16-23 is/are rejected under 35 U.S.C. 103 as being unpatentable over Rahman et al. (US-20230394235-A1), hereinafter Rahman, in view of Sethi et al. (US-20220374605-A1), hereinafter Sethi, further in view of Bonissone et al. (US-20150019269-A1), hereinafter Bonissone, further in view of Mukherjee et al. (US-12266431-B2), hereinafter Mukherjee.
Regarding claim 1, Rahman discloses: a method of utilizing case-based artificial intelligence reasoning ([0018] set of domain-specific rules, [In view of Fig. 6 of the instant application, the examiner is interpreting a case to be representative of one rule, i.e. one row of the case repository. Further, the examiner would like to note that the preamble of a claim does not necessarily provide patentable weight; therefore, the “artificial intelligence reasoning” does not require a mapping. Further still, Bonissone explicitly defines “a case-based reasoning system…different from other artificial intelligence approaches” (which also indicates Bonissone’s method to be an artificial intelligence approach for case-based reasoning), [0069].]) to convert natural language data (Abstract, natural language text) into constraints ([0022] findings that indicate whether the document meets a particular domain-specific requirement defined by a respective domain-specific rule [A constraint tracks to a requirement/rule]) via memory-based processing ([Fig. 3, Document Database 172, indicating the matching rules are memory-based, i.e. rules gathered from previous documents in memory]), the method being implemented by at least one processor ([0029] one or more processors), the method comprising:
receiving, by the at least one processor via a graphical user interface ([0058] generate graphical user interface data), at least one input ([0028] an end user may submit, using client device 104, a query via the web-based interface for documents [Web-based interface indicates a graphical representation]), each of the at least one input including input wording in a natural language format ([0030] the documents may include natural language prose, numbers, letters, and the like); and,
parsing, by the at least one processor using the at least one model ([0047] An NLP similarity recognition model), the at least one input to retrieve, from a case repository ([0047] identify sentences in a document that similar to one or more previously analyzed documents, [In view of document repository 202/document database 172 indicating retrieval from these storages for analysis]), at least one case that provides an existing solution to a first problem that is analogous to a second problem of the at least one input ([0046] template matching rules 500 may determine a similarity score indicating how similar a structure of text of a document is to a template structure, [0050] Entity-Value Matching: Entity-value matching rules 510 may include rules defining requirements that a particular entity or set of entities are represented by a document, have values resolvable to those entities, have valid values for those entities…for a document determined to be related to a first domain, entities 512 may include a set of entities expected to be within all documents related to the first domain, [Disclosure of domain-specific rules, wherein the rules may include template matching rules ([0045]) and/or entity-value matching rules ([0050]), further wherein a match is determined based on a similarity score ([0046]), indicates the rule being compared to new input for similarity is a solution to a first problem, i.e. document validation for that which generated the rule, that is analogous to a second problem, i.e. a new document where a similar rule is applied, e.g. to determine whether entities of the new document have resolvable values for purposes of validation (solving the problem), ([0050]), of the at least one input]), the retrieval including identification of the at least one case based on a predetermined similarity threshold ([0047] may identify similar documents, similar portions of documents (for example, similar sections, sentences, paragraphs, etc.), by computing a similarity metric, such as a cosine similarity, where, [0046] If the computed similarity score satisfies a threshold condition… [Indicating retrieval of documents based on a similarity threshold]).
Rahman does not disclose:
utilizing an artificial neural network technique to generate at least one model that comprises a Markov machine learning model of a randomly changing system.
Sethi discloses:
utilizing an artificial neural network technique to generate at least one model that comprises a Markov machine learning model of a randomly changing system ([0106] The models may include one or more of hidden Markov models, [0127] training samples may be based on one or more of random selection, [0147] FIG. 8 illustrates an example artificial neural network (“ANN”) 800. In particular embodiments, an ANN may refer to a computational model, [Disclosing an ANN which refers to a model while also disclosing that the model may contain Markov models indicates the ANN to be comprising a Markov machine learning model of a randomly changing system, i.e. during at least a training]).
Rahman, and Sethi are considered analogous art within document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman to incorporate the teachings of Sethi, because of the novel way to replace unknown entities in text with known entity names to improve the diversity and coverage of the training and testing samples (Sethi, [0134]).
Rahman in view of Sethi does not disclose:
utilizing the Markov machine learning model to adapt, by the at least one processor the retrieved at least one case to the at least one input.
Bonissone discloses:
utilizing the Markov machine learning model ([In view of the previously disclosed Markov machine learning model of Sethi]) to adapt, by the at least one processor ([In view of the previously disclosed processor of Rahman]) the retrieved at least one case to the at least one input ([0128] 5) Adapt the L refined solutions to the current case in order to derive a solution for the case, [Current case tracks to an input being adapted to retrieved cases]).
Rahman, Sethi, and Bonissone are considered analogous art within domain-specific document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman in view of Sethi to incorporate the teachings of Bonissone, because of the novel way to determine documents from previous cases relevant to a current case and adapting those solutions to the current case based on a confidence, improving quality of retrieved cases for adaptation (Bonissone, [0123]).
Rahman further discloses:
wherein the parsing includes determining at least one annotation for the at least one input ([0031] Tokenization process 210 may execute a process that converts a sequence of characters into a sequence of tokens, which may also be referred to as text tokens or lexical token [In view of the example provided in Fig. 6 of the instant application, annotation tracks to specific word/entity recognition. Tokenization of text tracks to identifying entities as tokens]), and wherein the at least one annotation includes a set of text abstractions and a corresponding word mapping ([0031] Each token may include a string of characters having a known meaning. The tokens may form an entity/value pair [Entity/value pair tracks to an abstraction, i.e. entity, and a corresponding word, i.e. value, mapping]).
Rahman in view of Sethi, further in view of Bonissone does not disclose:
wherein the at least one annotation includes a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept.
Mukherjee discloses:
wherein the at least one annotation includes a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept ([Fig. 10, Document Array (D)], [Col. 9, Lines 56-59] In various aspects, one or more of, or all of, the plurality of text blocks that were retrieved for the records sorted into the primary cluster are stored by the rule engine 104 in array “D”, [The symptoms/historical medical events gathered from the patient input of Fig. 7 being represented in an array, combined with the term/classification labels, i.e. the annotation, indicates the combination to be an appending of a numerical suffix, i.e. the classification, with a corresponding concept, i.e. the symptom/past medical history event, wherein each symptom having an associated classification, i.e. numeric suffix, indicates the numeric suffixes to be accounting for each occurrence of the concept (consider the two instances of classification “[R10.9]” in document array D). The examiner would like to note that a “numeric suffix that accounts for each occurrence of the corresponding concept” does not necessarily require an interpretation of that numeric suffix to be an overall count of total instances of the associated concepts, though the examiner believes this to be Applicant’s intent. Further, consider the “term frequency” metric of Mukherjee which suggest an overall count of classifications, [Col. 9, Lines 60-63].]),
Rahman, Sethi, Bonissone, and Mukherjee are considered analogous art within document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman in view of Sethi, further in view of Bonissone to incorporate the teachings of Mukherjee, because of the novel way to utilize patient contextual data and/or historical data in machine learning/rule-based diagnosis systems which improve in factual relevancy and contextual accuracy over time, enhancing user interaction with these systems (Mukherjee, [Col. 4, Lines 45-65]).
Sethi further discloses:
wherein the utilizing of the Markov machine learning model to adapt the retrieved at least one case to the at least one input comprises utilizing the Markov machine learning model ([In view of the previously disclosed Markov learning model of Sethi]) to compute, by the at least one processor ([0165] a processor 902), a merged mapping by:
updating, by the at least one processor ([In view of the previously disclosed processor of Sethi]), an input word mapping of the at least one input with a case word mapping of the retrieved at least one case ([0134] The extracted templates may be then provided to a data generator 536, which may further generate a plurality of synthetic dialog samples based on the one or more templates. For example, these synthetic dialog samples may include “what's the broadcast in New York,” “what's the broadcast in San Francisco,” etc. for the failure case of “what's the broadcast in Seattle?” [In view of the input and retrieved cases of Rahman, identifying a failure case, i.e. an unrecognized command, as input to be expanded to other templates with different people indicates the original “what’s the broadcast in Seattle?” as a retrieved case for inputs of “what’s the broadcast in New York?”, “what’s the broadcast in San Francisco?”, etc. where the entities, i.e. mappings, of the original and input are merged, i.e. updated, to determine the task of calling an individual]).
Rahman further discloses:
based on the retrieved at least one case ([Comparing a new document to a template indicates the similarity comparison for generating constraints (see below) is based upon the retrieved template case]), generating, by the at least one processor ([In view of the previously disclosed processor of Rahman/Sethi]), a constraint that comprises an if-then condition ([0046] if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule, [This tracks to an if: similarity score exceeds a threshold, then: classify as valid]).
Bonissone further discloses:
wherein generating the at least one constraint further comprises:
replacing, by the at least one processor ([In view of the previously disclosed processor of Sethi]), a set of text abstractions in the retrieved at least one case with information from the parsed at least one input by using the merged mapping ([Fig. 5, 508 “Adapt Retrieved Case to Current Case”], [In view of the entity merging of Sethi, adapting a retrieved case to a current case tracks to merging, i.e. replacing, the entities, i.e. text abstractions values, of the retrieved case with those parsed from the input case]).
Sethi further discloses:
generating, by the at least one processor, the at least one constraint by using a constraint template and a result of the replacing ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}”. This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training [Generating a constraint using a constraint template, i.e. please call {sl:contact}, and expanding to other people, i.e. “dad”, “Jill”, etc., indicates replacement of the original entity value “mom” (tracking to a retrieved case) with other known contacts for new constraints, i.e. “please call dad”]).
Rahman further discloses:
generating, by the at least one processor based on a result of the utilizing of the Markov machine learning model to adapt ([In view of the previously disclosed Markov machine learning model of Sethi]), at least one constraint that characterizes the at least one input, the at least one constraint relating to a rule that is mandated by the at least one input ([0049] In the template matching rule example, a document having a structure that is the same or similar to a template structure may produce a finding that the document complies with the template matching rule [Template matching “rule” tracks to a constraint to match pattern in order to satisfy compliance]).
Bonissone further discloses:
evaluating, by the at least one processor ([In view of the processor of Sethi/Rahman]), the at least one constraint ([0070] The retrieved, relevant cases are evaluated versus the current case, based on a confidence factor at step 510, [Applying the determined pattern/entity-value constraints of Rahman as the evaluation of confidence, i.e. confidence factor, in terms of the current case’s confidence/ability to fit the constraint/template/entity-value pattern of Rahman tracks to a constraint evaluation]).
Regarding claim 2, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the method of claim 1.
Rahman further discloses:
wherein the input wording includes at least one from among a word, a phrase, a sentence, a paragraph, and a document in the natural language format, the document including electronic data in a document file format ([0025] automatically validating unstructured documents having natural language text…Unstructured text is primarily composed of prose, and may include dates, numbers, and/or other forms of data. An example of unstructured data is unstructured text in journal articles [Journal articles are generally in the form of a document file (in view of the document database 172 indicating electronic data), in a natural language format, containing input of at least a word, phrase, sentence, and paragraph to convey the opinion of the author]).
Regarding claim 3, Rahman in view of Bonissone, further in view of Sethi, further in view of Mukherjee discloses: the method of claim 1.
Rahman further discloses:
determining, by the at least one processor using the at least one model ([In view of the previously disclosed NLP model and processor of Rahman]), a similarity value between the at least one input and each of a plurality of cases in the case repository ([0046] template matching rules 500 may determine a similarity score indicating how similar a structure of text of a document is to a template structure [In view of document database 172 indicating a plurality of cases to be compared to for similarity]), the similarity value relating to a distance between a plurality of data points in a similarity grouping ([0047] the similarity score, which is also referred to herein interchangeably as a similarity metric, refers to a distance between two feature vectors [Feature vectors indicate a plurality of data points, wherein those vectors are comprised of tokenized text]); and,
identifying, by the at least one processor ([In view of the previously disclosed processor of Rahman]), the at least one case based on the similarity value and the predetermined similarity threshold ([0046] If the computed similarity score satisfies a threshold condition, such as where the threshold condition is satisfied if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule [Wherein the template matching rules are gathered from the previously analyzed documents in document database 172 ([0047]) indicating identification of cases, i.e. previous documents, based on templates (rules) within those documents, using a similarity threshold]),
wherein the corresponding concept refers to an idea that is associated with each text abstraction from among the set of text abstractions ([0031] Each token may include a string of characters having a known meaning. The tokens may form an entity/value pair [Entity/value pair tracks to an abstraction, i.e. entity, and a corresponding word, i.e. value, mapping in view of the symptom/medical history of Mukherjee, wherein each symptom tracks to a corresponding concept of a previous medical event as would be abstracted using Rahman. Further, the examiner asserts that under BRI, the interpretation of a corresponding concept does not need to be different from the text abstraction itself. The examiner asserts that a text abstraction itself is a concept corresponding to an idea associated with text. Text abstractions, i.e. entity labels, as defined in Rahman will necessarily refer to ideas]).
Regarding claim 4, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the method of claim 3.
Rahman further discloses:
wherein determining the at least one annotation comprises:
accessing, by the at least one processor ([In view of the previously disclosed processor of Rahman]), at least one predefined list of concepts ([0018] a template matching rule, an entity-value matching rule [Template matching, i.e. specific word order rules (concepts), and entity-value matching, i.e. values expected to appear to satisfy rules (concepts)]), the concepts including a set of values that are expected to appear in the at least one input ([0019] The entity-value matching rule may require certain named entities extracted from the document are associated with values in the document that are allowed) and a text pattern matching expression that represents the set of values ([0018] the template matching rule may require that the content appear in a certain order in the document being validated [In view of the entity-value matching rules, a template matching rule containing content, i.e. entities, indicates a text pattern matching expression that represents the set of values, i.e. values of entities contained in a specific order within text documents]); and,
determining, by the at least one processor using the at least one model ([In view of the previously disclosed NLP model and processor of Rahman]), the at least one annotation for the at least one input based on the at least one predefined list of concepts ([0046] if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule. The similarity score may be computed by determining whether the structure of the text of the document is the same or similar to the example template structure [Identification of matching rules indicates entities, i.e. annotations, determined for the input based on a comparison to previously determined rules (lists of concepts), i.e. entity-value/template matching rules. In order to determine a match, there must be a knowledge of annotations in each document in order to make a comparison between the two documents]).
Regarding claim 7, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the method of claim 1.
Rahman further discloses:
presenting, by the at least one processor ([In view of the previously disclosed processor of Rahman]) via the graphical user interface ([In view of the previously disclosed GUI of Rahman]), a notification to at least one user associated with the at least one input ([Fig. 9A, 902a-n], [0062] an example user interface 900 may include section 902a, section 902b, ..., section 902n, each of which relates to a particular domain-specific rule that was evaluated against a document [Looking at figure 9A, the user will clearly be notified of the results in the form of the sections corresponding to rules 902a-n]), the notification including at least one from among the at least one constraint ([Fig. 9A, Edit Rule 916], [The ability to edit rules, i.e. constraints, indicates the user is notified to the constraint when editing]), a request for user feedback ([Fig. 9, Accept 908, Reject 910, Edit 912], [All forms of user feedback]), and information that relates to retrieval of the at least one case ([Fig. 9, Expected Result 904, Finding 906], [Findings of how well a rule fits a case tracks to information relating to retrieval of the case]);
determining, by the at least one processor ([In view of the previously disclosed processor of Rahman]), whether the generated at least one constraint includes information that corresponds to the at least one input based on the user feedback ([0062] Feedback 414 may include an indication of whether a particular finding was accepted, rejected, or edited [Wherein findings represent evaluations of domain specific rules, i.e. constraints, against documents ([0022]). Accepted findings indicate the generated constraint includes information that corresponds to the input. Rejected findings indicate the generated constraint does not include information that corresponds to the input.]),
wherein the user feedback is positive when the generated at least one constraint includes information that corresponds to the at least one input ([0022] As another example, feedback may indicate that a particular finding, corresponding to a particular domain- specific rule, is correct [“Correct” represents positive feedback regarding how a rule, i.e. constraint, corresponds to a document, i.e. input]); and,
wherein the user feedback is negative when the generated at least one constraint does not include information that corresponds to the at least one input ([0022] For example, feedback may indicate that a particular finding corresponding to a particular domain-specific rule is incorrect or should be updated [“Incorrect” represents negative feedback regarding how a rule, i.e. constraint, corresponds to a document, i.e. input]).
Regarding claim 8, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the method of claim 7.
Rahman further discloses:
requesting ([Fig. 9A, 912], [A button to edit a rule based on findings indicates a request to edit as having edit on each rule finding indicates the system “requests” the user to edit each finding, but the user is not required to edit]), by the at least one processor via the graphical user interface ([In view of the previously disclosed processor and graphical user interface of Rahman]), at least one correct constraint from the at least one user when the user feedback is negative ([0065] In some embodiments, rule updater 808 may update a rule based on feedback 414 [In view of the “reject” and “edit” buttons of Fig. 9A of Rahman, further in view of Fig. 9B indicating alternate correct structures that don’t match the original rule, i.e. resulting in corrections based on original negative user feedback]).
Sethi further discloses:
aggregating, by the at least one processor ([In view of the processor of Rahman]), data that corresponds to the at least one input ([0080] The NLU module 210 may further process information from these different sources by identifying and aggregating information [In view of Fig. 2 demonstrating the information is input through ASR system 208 resulting in text, in view of the input documents of Rahman. In view of the entity-value pairs of Rahman as the information gathered from NLU module 210]), the data including the input wording and at least one related annotation ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}” [{sl:contact} tracks to an annotation related to the input wording]);
computing, by the at least one processor ([In view of the previously disclosed processor of Rahman]), a new annotation for each of the at least one correct constraint ([0134] This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training [New annotations track to different values of the contact entity based on the constraint of them being after “call”, that the contact is known, etc.]); and,
generating, by the at least one processor ([In view of the previously disclosed processor of Rahman]), a new case by appending the new annotation to the aggregated data ([0134] The extracted templates may be then provided to a data generator 536, which may further generate a plurality of synthetic dialog samples based on the one or more templates. For example, these synthetic dialog samples may include “what's the broadcast in New York,” “what's the broadcast in San Francisco,” etc. for the failure case of “what's the broadcast in Seattle?” [Generating a plurality of samples, i.e. new cases, by appending a “location” annotation entity-value pair to aggregated broadcast data, in view of the previous aggregating NLU module 210]).
Rahman further discloses:
indexing, by the at least one processor ([In view of the previously disclosed processor of Rahman]), the new case for storage in the case repository ([0045] The set of domain-specific rules may be stored and/or access via rules database 174 [Domain-specific rules track to matching constraints in view of Rahman’s template matching rules]).
Regarding claim 9, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the method of claim 1.
Rahman further discloses:
wherein the at least one model includes at least one from among a natural language processing model ([0061] NLP model 616), a machine learning model ([0038] Machine learning models), a mathematical model ([Machine learning models are inherently statistically, i.e. mathematically, driven]), a process model ([0072] processing model), and a data model ([Fig. 1], [Transmitting/receiving data from a network 150 indicates the computing system 102 is a data model]).
Regarding claim 10, Rahman discloses: a computing device ([0026] computing system 102) configured to implement an execution of a method of utilizing case-based artificial intelligence reasoning ([0018] set of domain-specific rules, [In view of Fig. 6 of the instant application, the examiner is interpreting a case to be representative of one rule, i.e. one row of the case repository. Further, the examiner would like to note that the preamble of a claim does not necessarily provide patentable weight; therefore, the “artificial intelligence reasoning” does not require a mapping. Further still, Bonissone explicitly defines “a case-based reasoning system…different from other artificial intelligence approaches” (which also indicates Bonissone’s method to be an artificial intelligence approach for case-based reasoning), [0069].]) to convert natural language data (Abstract, natural language text) into constraints ([0022] findings that indicate whether the document meets a particular domain-specific requirement defined by a respective domain-specific rule [A constraint tracks to a requirement/rule]) via memory-based processing ([Fig. 3, Document Database 172, indicating the matching rules are memory-based, i.e. rules gathered from previous documents in memory]), the method being implemented by at least one processor ([0029] one or more processors), the computing device comprising:
a processor ([Fig. 12, Processor 1210-1]);
a memory ([Fig. 12, Memory 1220]); and
a communication interface coupled to each of the processor and the memory ([Fig. 12, I/O Interface 1250]), wherein the processor is configured to:
receive, via a graphical user interface ([0058] generate graphical user interface data), at least one input ([0028] an end user may submit, using client device 104, a query via the web-based interface for documents [Web-based interface indicates a graphical representation]), each of the at least one input including input wording in a natural language format ([0030] the documents may include natural language prose, numbers, letters, and the like); and,
parse, by using the at least one model ([0047] An NLP similarity recognition model), the at least one input to retrieve, from a case repository ([0047] identify sentences in a document that similar to one or more previously analyzed documents, [In view of document repository 202/document database 172 indicating retrieval from these storages for analysis]), at least one case that provides an existing solution to a first problem that is analogous to a second problem of the at least one input ([0046] template matching rules 500 may determine a similarity score indicating how similar a structure of text of a document is to a template structure, [0050] Entity-Value Matching: Entity-value matching rules 510 may include rules defining requirements that a particular entity or set of entities are represented by a document, have values resolvable to those entities, have valid values for those entities…for a document determined to be related to a first domain, entities 512 may include a set of entities expected to be within all documents related to the first domain, [Disclosure of domain-specific rules, wherein the rules may include template matching rules ([0045]) and/or entity-value matching rules ([0050]), further wherein a match is determined based on a similarity score ([0046]), indicates the rule being compared to new input for similarity is a solution to a first problem, i.e. document validation for that which generated the rule, that is analogous to a second problem, i.e. a new document where a similar rule is applied, e.g. to determine whether entities of the new document have resolvable values for purposes of validation (solving the problem), ([0050]), of the at least one input]), the retrieval including identification of the at least one case based on a predetermined similarity threshold ([0047] may identify similar documents, similar portions of documents (for example, similar sections, sentences, paragraphs, etc.), by computing a similarity metric, such as a cosine similarity, where, [0046] If the computed similarity score satisfies a threshold condition… [Indicating retrieval of documents based on a similarity threshold]).
Rahman does not disclose:
utilize an artificial neural network technique to generate at least one model that comprises a Markov machine learning model of a randomly changing system.
Sethi discloses:
utilize an artificial neural network technique to generate at least one model that comprises a Markov machine learning model of a randomly changing system ([0106] The models may include one or more of hidden Markov models, [0127] training samples may be based on one or more of random selection, [0147] FIG. 8 illustrates an example artificial neural network (“ANN”) 800. In particular embodiments, an ANN may refer to a computational model, [Disclosing an ANN which refers to a model while also disclosing that the model may contain Markov models indicates the ANN to be comprising a Markov machine learning model of a randomly changing system, i.e. during at least a training]).
Rahman, and Sethi are considered analogous art within document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman to incorporate the teachings of Sethi, because of the novel way to replace unknown entities in text with known entity names to improve the diversity and coverage of the training and testing samples (Sethi, [0134]).
Rahman in view of Sethi does not disclose:
utilize the Markov machine learning model to adapt the retrieved at least one case to the at least one input.
Bonissone discloses:
utilizing the Markov machine learning model ([In view of the previously disclosed Markov machine learning model of Sethi]) to adapt the retrieved at least one case to the at least one input ([0128] 5) Adapt the L refined solutions to the current case in order to derive a solution for the case, [Current case tracks to an input being adapted to retrieved cases]).
Rahman, Sethi, and Bonissone are considered analogous art within domain-specific document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman in view of Sethi to incorporate the teachings of Bonissone, because of the novel way to determine documents from previous cases relevant to a current case and adapting those solutions to the current case based on a confidence, improving quality of retrieved cases for adaptation (Bonissone, [0123]).
Rahman further discloses:
wherein the parse includes determining at least one annotation for the at least one input ([0031] Tokenization process 210 may execute a process that converts a sequence of characters into a sequence of tokens, which may also be referred to as text tokens or lexical token [In view of the example provided in Fig. 6 of the instant application, annotation tracks to specific word/entity recognition. Tokenization of text tracks to identifying entities as tokens]), and wherein the at least one annotation includes a set of text abstractions and a corresponding word mapping ([0031] Each token may include a string of characters having a known meaning. The tokens may form an entity/value pair [Entity/value pair tracks to an abstraction, i.e. entity, and a corresponding word, i.e. value, mapping]).
Rahman in view of Sethi, further in view of Bonissone does not disclose:
wherein the at least one annotation includes a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept.
Mukherjee discloses:
wherein the at least one annotation includes a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept ([Fig. 10, Document Array (D)], [Col. 9, Lines 56-59] In various aspects, one or more of, or all of, the plurality of text blocks that were retrieved for the records sorted into the primary cluster are stored by the rule engine 104 in array “D”, [The symptoms/historical medical events gathered from the patient input of Fig. 7 being represented in an array, combined with the term/classification labels, i.e. the annotation, indicates the combination to be an appending of a numerical suffix, i.e. the classification, with a corresponding concept, i.e. the symptom/past medical history event, wherein each symptom having an associated classification, i.e. numeric suffix, indicates the numeric suffixes to be accounting for each occurrence of the concept (consider the two instances of classification “[R10.9]” in document array D). The examiner would like to note that a “numeric suffix that accounts for each occurrence of the corresponding concept” does not necessarily require an interpretation of that numeric suffix to be an overall count of total instances of the associated concepts, though the examiner believes this to be Applicant’s intent. Further, consider the “term frequency” metric of Mukherjee which suggest an overall count of classifications, [Col. 9, Lines 60-63].]),
Rahman, Sethi, Bonissone, and Mukherjee are considered analogous art within document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman in view of Sethi, further in view of Bonissone to incorporate the teachings of Mukherjee, because of the novel way to utilize patient contextual data and/or historical data in machine learning/rule-based diagnosis systems which improve in factual relevancy and contextual accuracy over time, enhancing user interaction with these systems (Mukherjee, [Col. 4, Lines 45-65]).
Sethi further discloses:
wherein the processor is configured to utilize the Markov machine learning model ([In view of the previously disclosed Markov machine learning model of Sethi]) to adapt the retrieved at least one case to the at least one input by utilizing the Markov machine learning model to compute a merged mapping by:
updating an input word mapping of the at least one input with a case word mapping of the retrieved at least one case ([0134] The extracted templates may be then provided to a data generator 536, which may further generate a plurality of synthetic dialog samples based on the one or more templates. For example, these synthetic dialog samples may include “what's the broadcast in New York,” “what's the broadcast in San Francisco,” etc. for the failure case of “what's the broadcast in Seattle?” [In view of the input and retrieved cases of Rahman, identifying a failure case, i.e. an unrecognized command, as input to be expanded to other templates with different people indicates the original “what’s the broadcast in Seattle?” as a retrieved case for inputs of “what’s the broadcast in New York?”, “what’s the broadcast in San Francisco?”, etc. where the entities, i.e. mappings, of the original and input are merged, i.e. updated, to determine the task of calling an individual]).
Rahman further discloses:
based on the retrieved at least one case ([Comparing a new document to a template indicates the similarity comparison for generating constraints (see below) is based upon the retrieved template case]), generating a constraint that comprises an if-then condition ([0046] if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule, [This tracks to an if: similarity score exceeds a threshold, then: classify as valid]).
Bonissone further discloses:
wherein generating the at least one constraint further comprises:
replacing a set of text abstractions in the retrieved at least one case with information from the parsed at least one input by using the merged mapping ([Fig. 5, 508 “Adapt Retrieved Case to Current Case”], [In view of the entity merging of Sethi, adapting a retrieved case to a current case tracks to merging, i.e. replacing, the entities, i.e. text abstractions values, of the retrieved case with those parsed from the input case]).
Sethi further discloses:
generating the at least one constraint by using a constraint template and a result of the replacing ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}”. This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training [Generating a constraint using a constraint template, i.e. please call {sl:contact}, and expanding to other people, i.e. “dad”, “Jill”, etc., indicates replacement of the original entity value “mom” (tracking to a retrieved case) with other known contacts for new constraints, i.e. “please call dad”]).
Rahman further discloses:
generating based on a result of the utilizing of the Markov machine learning model to adapt ([In view of the previously disclosed Markov machine learning model of Sethi]), at least one constraint that characterizes the at least one input, the at least one constraint relating to a rule that is mandated by the at least one input ([0049] In the template matching rule example, a document having a structure that is the same or similar to a template structure may produce a finding that the document complies with the template matching rule [Template matching “rule” tracks to a constraint to match pattern in order to satisfy compliance]).
Bonissone further discloses:
evaluating the at least one constraint ([0070] The retrieved, relevant cases are evaluated versus the current case, based on a confidence factor at step 510, [Applying the determined pattern/entity-value constraints of Rahman as the evaluation of confidence, i.e. confidence factor, in terms of the current case’s confidence/ability to fit the constraint/template/entity-value pattern of Rahman tracks to a constraint evaluation]).
Regarding claim 11, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the computing device of claim 10.
Rahman further discloses:
wherein the input wording includes at least one from among a word, a phrase, a sentence, a paragraph, and a document in the natural language format, the document including electronic data in a document file format ([0025] automatically validating unstructured documents having natural language text…Unstructured text is primarily composed of prose, and may include dates, numbers, and/or other forms of data. An example of unstructured data is unstructured text in journal articles [Journal articles are generally in the form of a document file (in view of the document database 172 indicating electronic data), in a natural language format, containing input of at least a word, phrase, sentence, and paragraph to convey the opinion of the author]).
Regarding claim 12, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the computing device of claim 10.
Rahman further discloses:
determine, by using the at least one model ([In view of the previously disclosed NLP model of Rahman]), a similarity value between the at least one input and each of a plurality of cases in the case repository ([0046] template matching rules 500 may determine a similarity score indicating how similar a structure of text of a document is to a template structure [In view of document database 172 indicating a plurality of cases to be compared to for similarity]), the similarity value relating to a distance between a plurality of data points in a similarity grouping ([0047] the similarity score, which is also referred to herein interchangeably as a similarity metric, refers to a distance between two feature vectors [Feature vectors indicate a plurality of data points, wherein those vectors are comprised of tokenized text]); and,
identify, the at least one case based on the similarity value and the predetermined similarity threshold ([0046] If the computed similarity score satisfies a threshold condition, such as where the threshold condition is satisfied if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule [Wherein the template matching rules are gathered from the previously analyzed documents in document database 172 ([0047]) indicating identification of cases, i.e. previous documents, based on templates (rules) within those documents, using a similarity threshold]),
wherein the corresponding concept refers to an idea that is associated with each text abstraction from among the set of text abstractions ([0031] Each token may include a string of characters having a known meaning. The tokens may form an entity/value pair [Entity/value pair tracks to an abstraction, i.e. entity, and a corresponding word, i.e. value, mapping in view of the symptom/medical history of Mukherjee, wherein each symptom tracks to a corresponding concept of a previous medical event as would be abstracted using Rahman. Further, the examiner asserts that under BRI, the interpretation of a corresponding concept does not need to be different from the text abstraction itself. The examiner asserts that a text abstraction itself is a concept corresponding to an idea associated with text. Text abstractions, i.e. entity labels, as defined in Rahman will necessarily refer to ideas]).
Regarding claim 13, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the computing device of claim 12.
Rahman further discloses:
wherein, to determine the at least one annotation, the processor is configured to:
access at least one predefined list of concepts ([0018] a template matching rule, an entity-value matching rule [Template matching, i.e. specific word order rules (concepts), and entity-value matching, i.e. values expected to appear to satisfy rules (concepts)]), the concepts including a set of values that are expected to appear in the at least one input ([0019] The entity-value matching rule may require certain named entities extracted from the document are associated with values in the document that are allowed) and a text pattern matching expression that represents the set of values ([0018] the template matching rule may require that the content appear in a certain order in the document being validated [In view of the entity-value matching rules, a template matching rule containing content, i.e. entities, indicates a text pattern matching expression that represents the set of values, i.e. values of entities contained in a specific order within text documents]); and,
determine, by using the at least one model ([In view of the previously disclosed NLP model of Rahman]), the at least one annotation for the at least one input based on the at least one predefined list of concepts ([0046] if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule. The similarity score may be computed by determining whether the structure of the text of the document is the same or similar to the example template structure [Identification of matching rules indicates entities, i.e. annotations, determined for the input based on a comparison to previously determined rules (lists of concepts), i.e. entity-value/template matching rules. In order to determine a match, there must be a knowledge of annotations in each document in order to make a comparison between the two documents]).
Regarding claim 16, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the computing device of claim 10.
Rahman further discloses:
present, via the graphical user interface ([In view of the previously disclosed GUI of Rahman]), a notification to at least one user associated with the at least one input ([Fig. 9A, 902a-n], [0062] an example user interface 900 may include section 902a, section 902b, ..., section 902n, each of which relates to a particular domain-specific rule that was evaluated against a document [Looking at figure 9A, the user will clearly be notified of the results in the form of the sections corresponding to rules 902a-n]), the notification including at least one from among the at least one constraint ([Fig. 9A, Edit Rule 916], [The ability to edit rules, i.e. constraints, indicates the user is notified to the constraint when editing]), a request for user feedback ([Fig. 9, Accept 908, Reject 910, Edit 912], [All forms of user feedback]), and information that relates to retrieval of the at least one case ([Fig. 9, Expected Result 904, Finding 906], [Findings of how well a rule fits a case tracks to information relating to retrieval of the case]);
determine whether the generated at least one constraint includes information that corresponds to the at least one input based on the user feedback ([0062] Feedback 414 may include an indication of whether a particular finding was accepted, rejected, or edited [Wherein findings represent evaluations of domain specific rules, i.e. constraints, against documents ([0022]). Accepted findings indicate the generated constraint includes information that corresponds to the input. Rejected findings indicate the generated constraint does not include information that corresponds to the input.]),
wherein the user feedback is positive when the generated at least one constraint includes information that corresponds to the at least one input ([0022] As another example, feedback may indicate that a particular finding, corresponding to a particular domain- specific rule, is correct [“Correct” represents positive feedback regarding how a rule, i.e. constraint, corresponds to a document, i.e. input]); and,
wherein the user feedback is negative when the generated at least one constraint does not include information that corresponds to the at least one input ([0022] For example, feedback may indicate that a particular finding corresponding to a particular domain-specific rule is incorrect or should be updated [“Incorrect” represents negative feedback regarding how a rule, i.e. constraint, corresponds to a document, i.e. input]).
Regarding claim 17, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the computing device of claim 16.
Rahman further discloses:
request ([Fig. 9A, 912], [A button to edit a rule based on findings indicates a request to edit as having edit on each rule finding indicates the system “requests” the user to edit each finding, but the user is not required to edit]), via the graphical user interface ([In view of the previously disclosed graphical user interface of Rahman]), at least one correct constraint from the at least one user when the user feedback is negative ([0065] In some embodiments, rule updater 808 may update a rule based on feedback 414 [In view of the “reject” and “edit” buttons of Fig. 9A of Rahman, further in view of Fig. 9B indicating alternate correct structures that don’t match the original rule, i.e. resulting in corrections based on original negative user feedback]);
Sethi further discloses:
aggregate data that corresponds to the at least one input ([0080] The NLU module 210 may further process information from these different sources by identifying and aggregating information [In view of Fig. 2 demonstrating the information is input through ASR system 208 resulting in text, in view of the input documents of Rahman. In view of the entity-value pairs of Rahman as the information gathered from NLU module 210]), the data including the input wording and at least one related annotation ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}” [{sl:contact} tracks to an annotation related to the input wording]);
compute a new annotation for each of the at least one correct constraint ([0134] This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training [New annotations track to different values of the contact entity based on the constraint of them being after “call”, that the contact is known, etc.]); and,
generate a new case by appending the new annotation to the aggregated data ([0134] The extracted templates may be then provided to a data generator 536, which may further generate a plurality of synthetic dialog samples based on the one or more templates. For example, these synthetic dialog samples may include “what's the broadcast in New York,” “what's the broadcast in San Francisco,” etc. for the failure case of “what's the broadcast in Seattle?” [Generating a plurality of samples, i.e. new cases, by appending a “location” annotation entity-value pair to aggregated broadcast data, in view of the previous aggregating NLU module 210]).
Rahman further discloses:
index the new case for storage in the case repository ([0045] The set of domain-specific rules may be stored and/or access via rules database 174 [Domain-specific rules track to matching constraints in view of Rahman’s template matching rules]).
Regarding claim 18, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the computing device of claim 10.
Rahman further discloses:
wherein the at least one model includes at least one from among a natural language processing model ([0061] NLP model 616), a machine learning model ([0038] Machine learning models), a mathematical model ([Machine learning models are inherently statistically, i.e. mathematically, driven]), a process model ([0072] processing model), and a data model ([Fig. 1], [Transmitting/receiving data from a network 150 indicates the computing system 102 is a data model]).
Regarding claim 19, Rahman discloses: a non-transitory computer readable storage medium ([0090] non-transitory computer readable storage medium) storing instructions that utilize case-based artificial intelligence reasoning ([0018] set of domain-specific rules, [In view of Fig. 6 of the instant application, the examiner is interpreting a case to be representative of one rule, i.e. one row of the case repository. Further, the examiner would like to note that the preamble of a claim does not necessarily provide patentable weight; therefore, the “artificial intelligence reasoning” does not require a mapping. Further still, Bonissone explicitly defines “a case-based reasoning system…different from other artificial intelligence approaches” (which also indicates Bonissone’s method to be an artificial intelligence approach for case-based reasoning), [0069].]) to convert natural language data (Abstract, natural language text) into constraints ([0022] findings that indicate whether the document meets a particular domain-specific requirement defined by a respective domain-specific rule [A constraint tracks to a requirement/rule]) via memory-based processing, the storage medium comprising executable code which ([0090] Instructions or other program code), when executed by a processor ([Fig. 12, Processor 1210-1]), causes the processor to:
receive, via a graphical user interface ([0058] generate graphical user interface data), at least one input ([0028] an end user may submit, using client device 104, a query via the web-based interface for documents [Web-based interface indicates a graphical representation]), each of the at least one input including input wording in a natural language format ([0030] the documents may include natural language prose, numbers, letters, and the like); and,
parse, by using the at least one model ([0047] An NLP similarity recognition model), the at least one input to retrieve, from a case repository ([0047] identify sentences in a document that similar to one or more previously analyzed documents, [In view of document repository 202/document database 172 indicating retrieval from these storages for analysis]), at least one case that provides an existing solution to a first problem that is analogous to a second problem of the at least one input ([0046] template matching rules 500 may determine a similarity score indicating how similar a structure of text of a document is to a template structure, [0050] Entity-Value Matching: Entity-value matching rules 510 may include rules defining requirements that a particular entity or set of entities are represented by a document, have values resolvable to those entities, have valid values for those entities…for a document determined to be related to a first domain, entities 512 may include a set of entities expected to be within all documents related to the first domain, [Disclosure of domain-specific rules, wherein the rules may include template matching rules ([0045]) and/or entity-value matching rules ([0050]), further wherein a match is determined based on a similarity score ([0046]), indicates the rule being compared to new input for similarity is a solution to a first problem, i.e. document validation for that which generated the rule, that is analogous to a second problem, i.e. a new document where a similar rule is applied, e.g. to determine whether entities of the new document have resolvable values for purposes of validation (solving the problem), ([0050]), of the at least one input]), the retrieval including identification of the at least one case based on a predetermined similarity threshold ([0047] may identify similar documents, similar portions of documents (for example, similar sections, sentences, paragraphs, etc.), by computing a similarity metric, such as a cosine similarity, where, [0046] If the computed similarity score satisfies a threshold condition… [Indicating retrieval of documents based on a similarity threshold]).
Rahman does not disclose:
utilize an artificial neural network technique to generate at least one model that comprises a Markov machine learning model of a randomly changing system.
Sethi discloses:
utilize an artificial neural network technique to generate at least one model that comprises a Markov machine learning model of a randomly changing system ([0106] The models may include one or more of hidden Markov models, [0127] training samples may be based on one or more of random selection, [0147] FIG. 8 illustrates an example artificial neural network (“ANN”) 800. In particular embodiments, an ANN may refer to a computational model, [Disclosing an ANN which refers to a model while also disclosing that the model may contain Markov models indicates the ANN to be comprising a Markov machine learning model of a randomly changing system, i.e. during at least a training]).
Rahman, and Sethi are considered analogous art within document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman to incorporate the teachings of Sethi, because of the novel way to replace unknown entities in text with known entity names to improve the diversity and coverage of the training and testing samples (Sethi, [0134]).
Rahman in view of Sethi does not disclose:
utilize the Markov machine learning model to adapt the retrieved at least one case to the at least one input.
Bonissone discloses:
utilizing the Markov machine learning model ([In view of the previously disclosed Markov machine learning model of Sethi]) to adapt the retrieved at least one case to the at least one input ([0128] 5) Adapt the L refined solutions to the current case in order to derive a solution for the case, [Current case tracks to an input being adapted to retrieved cases]).
Rahman, Sethi, and Bonissone are considered analogous art within domain-specific document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman in view of Sethi to incorporate the teachings of Bonissone, because of the novel way to determine documents from previous cases relevant to a current case and adapting those solutions to the current case based on a confidence, improving quality of retrieved cases for adaptation (Bonissone, [0123]).
Rahman further discloses:
wherein the parse includes determining at least one annotation for the at least one input ([0031] Tokenization process 210 may execute a process that converts a sequence of characters into a sequence of tokens, which may also be referred to as text tokens or lexical token [In view of the example provided in Fig. 6 of the instant application, annotation tracks to specific word/entity recognition. Tokenization of text tracks to identifying entities as tokens]), and wherein the at least one annotation includes a set of text abstractions and a corresponding word mapping ([0031] Each token may include a string of characters having a known meaning. The tokens may form an entity/value pair [Entity/value pair tracks to an abstraction, i.e. entity, and a corresponding word, i.e. value, mapping]).
Rahman in view of Sethi, further in view of Bonissone does not disclose:
wherein the at least one annotation includes a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept.
Mukherjee discloses:
wherein the at least one annotation includes a corresponding concept that is appended to a numeric suffix that accounts for each occurrence of the corresponding concept ([Fig. 10, Document Array (D)], [Col. 9, Lines 56-59] In various aspects, one or more of, or all of, the plurality of text blocks that were retrieved for the records sorted into the primary cluster are stored by the rule engine 104 in array “D”, [The symptoms/historical medical events gathered from the patient input of Fig. 7 being represented in an array, combined with the term/classification labels, i.e. the annotation, indicates the combination to be an appending of a numerical suffix, i.e. the classification, with a corresponding concept, i.e. the symptom/past medical history event, wherein each symptom having an associated classification, i.e. numeric suffix, indicates the numeric suffixes to be accounting for each occurrence of the concept (consider the two instances of classification “[R10.9]” in document array D). The examiner would like to note that a “numeric suffix that accounts for each occurrence of the corresponding concept” does not necessarily require an interpretation of that numeric suffix to be an overall count of total instances of the associated concepts, though the examiner believes this to be Applicant’s intent. Further, consider the “term frequency” metric of Mukherjee which suggest an overall count of classifications, [Col. 9, Lines 60-63].]),
Rahman, Sethi, Bonissone, and Mukherjee are considered analogous art within document analysis. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Rahman in view of Sethi, further in view of Bonissone to incorporate the teachings of Mukherjee, because of the novel way to utilize patient contextual data and/or historical data in machine learning/rule-based diagnosis systems which improve in factual relevancy and contextual accuracy over time, enhancing user interaction with these systems (Mukherjee, [Col. 4, Lines 45-65]).
Sethi further discloses:
wherein the processor is configured to utilize the Markov machine learning model ([In view of the previously disclosed Markov machine learning model of Sethi]) to adapt the retrieved at least one case to the at least one input by utilizing the Markov machine learning model to compute a merged mapping by:
updating an input word mapping of the at least one input with a case word mapping of the retrieved at least one case ([0134] The extracted templates may be then provided to a data generator 536, which may further generate a plurality of synthetic dialog samples based on the one or more templates. For example, these synthetic dialog samples may include “what's the broadcast in New York,” “what's the broadcast in San Francisco,” etc. for the failure case of “what's the broadcast in Seattle?” [In view of the input and retrieved cases of Rahman, identifying a failure case, i.e. an unrecognized command, as input to be expanded to other templates with different people indicates the original “what’s the broadcast in Seattle?” as a retrieved case for inputs of “what’s the broadcast in New York?”, “what’s the broadcast in San Francisco?”, etc. where the entities, i.e. mappings, of the original and input are merged, i.e. updated, to determine the task of calling an individual]).
Rahman further discloses:
based on the retrieved at least one case ([Comparing a new document to a template indicates the similarity comparison for generating constraints (see below) is based upon the retrieved template case]), generating a constraint that comprises an if-then condition ([0046] if the computed similarity score is equal to or greater than a threshold similarity score, then that document may be classified as valid against the evaluation of the template matching rule, [This tracks to an if: similarity score exceeds a threshold, then: classify as valid]).
Bonissone further discloses:
wherein generating the at least one constraint further comprises:
replacing a set of text abstractions in the retrieved at least one case with information from the parsed at least one input by using the merged mapping ([Fig. 5, 508 “Adapt Retrieved Case to Current Case”], [In view of the entity merging of Sethi, adapting a retrieved case to a current case tracks to merging, i.e. replacing, the entities, i.e. text abstractions values, of the retrieved case with those parsed from the input case]).
Sethi further discloses:
generating the at least one constraint by using a constraint template and a result of the replacing ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}”. This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training [Generating a constraint using a constraint template, i.e. please call {sl:contact}, and expanding to other people, i.e. “dad”, “Jill”, etc., indicates replacement of the original entity value “mom” (tracking to a retrieved case) with other known contacts for new constraints, i.e. “please call dad”]).
Rahman further discloses:
generating based on a result of the utilizing of the Markov machine learning model to adapt ([In view of the previously disclosed Markov machine learning model of Sethi]), at least one constraint that characterizes the at least one input, the at least one constraint relating to a rule that is mandated by the at least one input ([0049] In the template matching rule example, a document having a structure that is the same or similar to a template structure may produce a finding that the document complies with the template matching rule [Template matching “rule” tracks to a constraint to match pattern in order to satisfy compliance]).
Bonissone further discloses:
evaluating the at least one constraint ([0070] The retrieved, relevant cases are evaluated versus the current case, based on a confidence factor at step 510, [Applying the determined pattern/entity-value constraints of Rahman as the evaluation of confidence, i.e. confidence factor, in terms of the current case’s confidence/ability to fit the constraint/template/entity-value pattern of Rahman tracks to a constraint evaluation]).
Regarding claim 20, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the storage medium of claim 19.
Rahman further discloses:
wherein the input wording includes at least one from among a word, a phrase, a sentence, a paragraph, and a document in the natural language format, the document including electronic data in a document file format ([0025] automatically validating unstructured documents having natural language text…Unstructured text is primarily composed of prose, and may include dates, numbers, and/or other forms of data. An example of unstructured data is unstructured text in journal articles [Journal articles are generally in the form of a document file (in view of the document database 172 indicating electronic data), in a natural language format, containing input of at least a word, phrase, sentence, and paragraph to convey the opinion of the author]).
Regarding claim 21, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the method of claim 1.
Sethi further discloses:
wherein the generating the at least one constraint, comprises utilizing memory-based processing to translate natural language data into the at least one constraint ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}”. This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training, [Generation of a constraint, i.e. “please call dad/Jill”, based on a previously generated constraint, i.e. “please call mom”, indicating “please call mom” to be a memory-based processing example for generation of the extended training samples into constraints based on natural language for calling dad, Jill, etc.]).
Regarding claim 22, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the computing device of claim 10.
Sethi further discloses:
wherein the generating the at least one constraint, comprises utilizing memory-based processing to translate natural language data into the at least one constraint ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}”. This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training, [Generation of a constraint, i.e. “please call dad/Jill”, based on a previously generated constraint, i.e. “please call mom”, indicating “please call mom” to be a memory-based processing example for generation of the extended training samples into constraints based on natural language for calling dad, Jill, etc.]).
Regarding claim 23, Rahman in view of Sethi, further in view of Bonissone, further in view of Mukherjee discloses: the storage medium of claim 19.
Sethi further discloses:
wherein the generating the at least one constraint, comprises utilizing memory-based processing to translate natural language data into the at least one constraint ([0134] For example, if “please call mom” is a failure case, the template extractor 534 may extract a template as “please call {sl:contact}”. This template may then be expanded to other similar utterances such as “please call dad”, “please call Jill”, etc. and added to training, [Generation of an extended template constraint, i.e. “please call dad/Jill”, based on a previously generated constraint, i.e. “please call mom”, indicating “please call mom” to be a memory-based processing example for generation of the extended training samples into constraints based on natural language for calling dad, Jill, etc.]).
Conclusion
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.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Lopez Garcia et al. (US-20230297784-A1) discloses “a method for automated decision modelling from text including obtaining a text corpus including a policy. Terms and syntax are identified within the text corpus related to the policy. Sentence similarities and co-references based on the terms and syntax are identified. Discourse and sentence level semantic parsing is performed based on the terms and the sentence similarities and the co-references using machine learning. A decision model template is generated based on the discourse and semantic parsing, and the decision model template is transformed into an automated decision model” (abstract). Specifically, Lopez Garcia discloses annotating similar sentences using labels which include numbers for purposes of determining repeated and/or matching rules/sentences ([0040]). See entire document.
Rogynskyy et al. (US-20250045313-A1) discloses “systems and methods for automatic generation of datasets for record objects using one or more large language models. A system can identify a plurality of electronic activities. The system can generate, using one or more large language models, a first set of text strings from the electronic activities. The system can store, in one or more data structures, a first association between the first set of text strings and the record object. The system can identify a second electronic activity. The system can generate, using the one or more large language models, a second set of text strings based on data corresponding to the second electronic activity and the first set of text strings. The system can store, in the one or more data structures, a second association between the second set of text strings and the record object” (abstract). Specifically, there is disclosure of a counter for tracking the amount of times an entity appears within text strings ([0447]). The examiner would like to note that this document does not beat the EFD of the instant application. It has been cited to aid in describing the state of the art.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to THEODORE JOHN WITHEY whose telephone number is (703)756-1754. The examiner can normally be reached Monday - Friday, 8am-5pm.
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, Andrew Flanders can be reached on (571) 272-7516. 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.
/THEODORE WITHEY/Examiner, Art Unit 2655
/ANDREW C FLANDERS/Supervisory Patent Examiner, Art Unit 2655
Prosecution Insights