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 .
DETAILED ACTION
Claims 1-15, 17-20 are presented for examination.
Claims 1, 4, 8, 11, 14and 17, 18 and 20 were amended.
Claim 16 is cancelled.
This is a Non-Final Action.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 06/09/2026 has been entered.
Applicant’s arguments with respect to claim(s) 1-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
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 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 of this title, 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 1-4, 8-14 and 17-21 are rejected under 35 U.S.C. 103 as being unpatentable over Chowdhury et al. (US 11,138,523) in view of Gunda et al. (US 2023/0028278) further in view of Sachdev et al. (US 11,386,365) and Juneja et al. (US 2021/0304206)
1. Chowdhury teaches, A computer-implemented method comprising:
generating, by one or more processors, a target object embedding (Col 18: lines 30-31 - teaches use LSA distributed semantic model to generate LSA similarity score – discloses the LSA semantic model generates vector representations (semantic embeddings) used to compute similarity between objects, Chowdhury) for an unlabeled data object within a set of data objects that comprise a subset of unlabeled data objects and a subset of labeled data objects (Col 20: lines 25-48 & Fig 5: 504 – teaches receive non-annotated corpus of source input… 508 – teaches repository of unannotated instances and seeds, 512 – teaches repository of annotated seeds – disclosing a dataset including unannotated (unlabeled) and annotated (labeled) instances used by the active learning system, Chowdhury), wherein a labeled data object of the subset of labeled data objects is assigned (a) a binary classification label (Abstract - teaches the first and second instances are respectively annotated with a positive and negative label – discloses Positive/Negative labels corresponds to a binary classification label assigned to labeled data objects, Chowdhury)
generating, by the one or more processors, an embedding similarity score for the unlabeled data object based on at least one of an angle or a distance between the target object embedding and a reference object embedding corresponding to the labeled data object (Col 13: lines 8-16 – teaches SVM builds a model representing examples as points in space, mapped such that examples of the two categories are separated by as wide a gap; new examples are then mapped into the same model space, and according to which side of the gap they fall, a prediction is made as to which category they are associated with – disclose determining unlabeled objects using embedding similarity scores and quantile threshold tied to labeled object rating, Chowdhury);
determining, by the one or more processors, the unlabeled data object from the subset of unlabeled data objects based on the similarity score (Col 18: lines 18-22 – teaches using LSA similarity scores … to rank unannotated instances – disclosing computation of similarity scores between vector representations of instances, Chowdhury); and
…that comprises the unlabeled data object (Col 17: lines 54-55 - teaches provides the selected unannotated candidate seed to the user and alternatively provides relevant source input including unannotated instances to the user in ranked order, Chowdhury).
Chowdhury does not explicitly teach,
(b) a quantile rating of set of quantile ratings that identifies a position of the labeled data object relative to the subset of labeled data objects, (ii) the quantile rating corresponds to a quantile ranking of a set of quantile rankings, (iii) the quantile ranking is derived from a numerical attribute of the labeled data object;
… a quantile similarity threshold that corresponds to the quantile rating of the labeled data object
providing, by the one or more processors, a message.
However, Gunda teaches,
(b) a quantile rating of set of quantile ratings that identifies a position of the labeled data object relative to the subset of labeled data objects (Paragraph 89 – teaches the quantile function may return a particular quantile (e.g. decile, percentile, etc.) or organizational data 311… in comparison to other corresponding organizational data 311, permitting “4” quartiles, “5” quintiles, “10” deciles, or “100” percentiles, Gunda), (ii) the quantile rating corresponds to a quantile ranking of a set of quantile rankings (Paragraph 91-92 – teaches Denise smith may fall into any one of 10 buckets depending on how her ‘compensation’ ranked in the top ten percent, “10” is returned as the decile bucket, whereas if it ranked last, “1” is returned, Gunda), (iii) the quantile ranking is derived from a numerical attribute of the labeled data object (Paragraphs 89 and 91 – teaches the second example argument of the ‘quantile’ function may specify a numeric attribute value … for comparison then identifies compensation as the attribute of the specified employee to be compared, Gunda).
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to allow apply Gunda’s quantile-based characterization of numerical object attributes to Chowdhury’s labeled instances to provide an additional standardized indication of the relative position of each labeled instance within a comparison population. Gunda expressly teaches deriving such a quantile from a numerical attribute of the particular object and returning a corresponding relative-ranking bucket.
However, Sachdev teaches,
… a quantile similarity threshold that corresponds to the quantile rating of the labeled data object (Col 7: lines 35-44 - calculate and/or update quantile thresholds from distributions of scores for the candidates – disclosing quantile threshold used to determine candidate ranking positions, Sachdev).
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to apply Sachdev’s quantile-threshold technique to the numerical similarity scores generated by Chowdhury because Sachdev teaches using score values associated with selected quantiles as thresholds to efficiently determine relative positions without processing the complete ranking. This would predictably permit Chowdhury’s candidate selection process to use thresholds corresponding to the quantile information supplied according to the Gunda.
However, Juneja teaches,
providing, by the one or more processors, a message (Paragraph 7 – teaches determine at least one authorization action for processing the transaction request – disclosing providing a system output/response based on machine learning scoring and threshold evaluations, Juneja).
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to allow Chowdhury to be combined with Juneja because both Chowdhury and Juneja are in the same field of endeavor of machine learning system that manage data selection and prioritization and by combining the prior arts would allow for delivering instance in a structured, prioritized expansion message.
2. The combination of Chowdhury, Gunda, Sachdev and Juneja teach, The computer-implemented method of claim 1, further comprising:
generating, using a binary classification model, a positive classification prediction for an input data object of the set of data objects (Paragraph 29 – teaches generate a recovery score based on the transaction data and merchant data associated with the merchant, the recovery score representing a likelihood that an issuer system associated with the account of the user will be able to recover at least a portion of the payment amount if the payment transaction is reversed; and Paragraph 62 – teaches at step 304, a fraud risk score is generated. The fraud risk score represents a likelihood that the requested transaction is fraudulent (i.e. binary, fraudulent vs not fraudulent), the transaction represents the input data, Juneja);
receiving a recovery value for the input data object (Paragraph 64 - at step 306, a recovery score is generated. The recovery score represents a likelihood that the merchant and/or issuer will recover the transaction value or portion thereof if the transaction is fraudulent and/or reversed, disputed and/or the like, Juneja); and
generating the labeled data object by assigning the binary classification label and the quantile rating to the input data object based on the recovery value (Paragraph 29 – clause 17 - teaches “generate a recovery score… determine at least one authorization action…; and process the transaction – disclosing that decisions/actions are made based on both recovery and fraud score thereby corresponds to assigning labels and prioritization (quantile style_ to data objects, Juneja).
3. The combination of Chowdhury, Gunda, Sachdev and Juneja teach, The computer-implemented method of claim 2, wherein the subset of labeled data objects is respectively associated with a set of recovery values and assigning the quantile rating to the input data object (Paragraph 64 - at step 306, a recovery score is generated. The recovery score represents a likelihood that the merchant and/or issuer will recover the transaction value or portion thereof if the transaction is fraudulent and/or reversed, disputed and/or the like, Juneja) comprises:
generating a set of quantile rankings based on the set of recovery values (Fig 4, Paragraph 69 – teaches now to FIG. 4, shown are two recovery score determinations according to non-limiting embodiments; In each example, a set of input parameters 402, 404 result in a respective recovery score 406, 408. In this example, the scores are 7 and 95 – disclosing values naturally lends themselves to ranking/quantile partitioning, Juneja);
determining a quantile ranking for the input data object from the set of quantile rankings based on the recovery value (Paragraph 69 - Input parameters 402 result in a low recovery score (7); and input parameters 404 result in a high recovery score (95) -disclosing low vs high recovery scores); and
assigning the quantile rating based on the quantile ranking (paragraph 7 teaches - determine at least one authorization action for processing the transaction request based on the recovery score and the fraud risk score; and process the transaction request based on the at least one authorization action, Juneja).
4. The combination of Chowdhury, Gunda, Sachdev and Juneja teach, The computer-implemented method of claim 3, wherein the quantile rating is one of a set of quantile ratings (Col 6: lines 50-65 – teaches quantile thresholds 216 may include score values that denote fixed percentiles, such as the top 0%, 10%, 20%, 50% and 100% of scores 226, further teaching that ordered scores and their relative positions are used to determine the corresponding quantile threshold, Sachdev), the quantile similarity threshold is one of a set of quantile similarity thresholds, and the set of quantile similarity thresholds respectively corresponds to the set of quantile rankings (Col 7: lines 25-33 - teaches the system orders scores and uses selected quantiles to determine the corresponding quantile threshold, Sachdev), and a magnitude of the quantile similarity threshold is relative to the quantile rating(Col 8: lines 1-17 – teaches an array of scores… corresponding to percentiles and expressly associates score values 2, 3, 4,5, 8, 9, 13 with percentiles 0, 5, 10, 20, 50, 100, Sachdev).
8. The combination of Chowdhury, Gunda, Sachdev and Juneja teach, The computer-implemented method of claim 1, wherein the message comprises an expansion response to an expansion request (Paragraph 7 – teaches generate a recovery score based on the transaction data and merchant data associated with the merchant, the recovery score representing a likelihood that an issuer system associated with the account of the user will be able to recover at least a portion of the payment amount if the payment transaction is reversed; determine at least one authorization action for processing the transaction request based on the recovery score and the fraud risk score; and process the transaction request based on the at least one authorization action – discloses a system that responds to incoming transaction (expansion request) with an authorization action (expansion response) based on fraud and recovery scores , Juneja) and the computer-implemented method further comprises:
receiving a scenario identifier (Paragraph 54 - The historical transaction data may include previous transactions with the same account identifier, with the same merchant, in the same merchant category, within a geographical location, and/or the like – discloses identifiers such as merchant ID, merchant category code or account ID, which act as scenario identifiers to contextualize transactions, Juneja);
determining an initial subset of target data objects from the set of data objects based on the scenario identifier (Paragraph 64 – The recovery score may be based on transaction parameters from the requested transaction according to one or more models generated based on historical transaction data. The models may be machine learning models. The historical transaction data may include previous transactions with the same account identifier, with the same merchant, in the same merchant category, within a geographical location – disclose filtering transactions (data objects) by scenario identifiers such as merchant category or account ID, forming an initial subset, Juneja);
generating the subset of labeled data objects by determining a respective quantile rating associated with up to each of the initial subset of target data objects (Paragraph 69 – teaches Referring now to FIG. 4, shown are two recovery score determinations according to non-limiting embodiments. In each example, a set of input parameters 402, 404 result in a respective recovery score 406, 408. In this example, the scores are 7 and 95. Input parameters 402 result in a low recovery score (7) (e.g., a low risk of not being able to recover) due in part to the merchant country code (840 corresponding to the United States), relatively low transaction value ($200), and participation in fraud deflection services. Each of these parameters and others may be weighed to calculate the recovery score. Input parameters 404 result in a high recovery score (95) – discloses numerical recovery scores that can be bucketed into quantile ratings (low vs high); these rating can be assigned to the initial subset of objects, Juneja); and
in response to the expansion request, determining an expanded subset of target data objects from the set of data objects based on the initial subset of target data objects, wherein the expanded subset of target data objects comprises the unlabeled data object (Paragraph 7 - generate a recovery score based on the transaction data and merchant data associated with the merchant, the recovery score representing a likelihood that an issuer system associated with the account of the user will be able to recover at least a portion of the payment amount if the payment transaction is reversed; determine at least one authorization action for processing the transaction request based on the recovery score and the fraud risk score – discloses dynamic decision making where more objects (transactions) can be evaluated based on prior subsets and scores; this corresponds to expanding the initial subset to include additional (previously unlabeled) objects in response to a request, Juneja).
9. The combination of Chowdhury, Gunda, Sachdev and Juneja teach, The computer-implemented method of claim 8, wherein the expanded subset of target data objects comprises a ranking of unlabeled data objects from the subset of unlabeled data objects Paragraph 69 – teaches Referring now to FIG. 4, shown are two recovery score determinations according to non-limiting embodiments. In each example, a set of input parameters 402, 404 result in a respective recovery score 406, 408. In this example, the scores are 7 and 95. Input parameters 402 result in a low recovery score (7) (e.g., a low risk of not being able to recover) due in part to the merchant country code (840 corresponding to the United States), relatively low transaction value ($200), and participation in fraud deflection services. Each of these parameters and others may be weighed to calculate the recovery score. Input parameters 404 result in a high recovery score (95) – discloses that recovery scores can be high or low, providing a natural ordering of objects. This corresponds to ranking data objects within an expanded subset, Juneja) and the unlabeled data object is positioned within the ranking based on the quantile rating of the labeled data object (Paragraph 7 - generate a recovery score based on the transaction data and merchant data associated with the merchant, the recovery score representing a likelihood that an issuer system associated with the account of the user will be able to recover at least a portion of the payment amount if the payment transaction is reversed; determine at least one authorization action for processing the transaction request based on the recovery score and the fraud risk score – disclose decision making where actions are determined based on recovery and fraud scores. When applied to multiple transactions, these scores position objects relative to each other (low recovery vs high recovery, Juneja).
10. The combination of Chowdhury, Gunda, Sachdev and Juneja teach, The computer-implemented method of claim 1, further comprising: receiving a verification input that comprise a recovery value for the unlabeled data object (Paragraph 64 - at step 306, a recovery score is generated. The recovery score represents a likelihood that the merchant and/or issuer will recover the transaction value or a portion thereof if the transaction is fraudulent and/or reversed, disputed, and/or the like – discloses that recovery scores (recovery values) are received/generated for transactions this is similar to receiving a verification input comprising a recovery value, Juneja); and
in response to the verification input, (i) assigning (a) a new binary classification label (Paragraph 62, Fig 3: 304 – teaches a fraud risk score is generated. The fraud risk score represents a likelihood that the requested transaction is fraudulent. The fraud risk score may be generated based on transaction parameters from the requested transaction according to one or more models generated based on historical transaction data, Juneja)and (b) a new quantile rating to the unlabeled data object and (ii) storing the unlabeled data object as a new labeled data object (Abstract teaches - generating a recovery score; determining at least one authorization action for processing the transaction request based on the recovery score and the fraud risk score; and processing the transaction request based on the at least one authorization action, Juneja).
21. The combination of Chowdhury, Gunda, Sachdev and Juneja teach, The computer-implemented method of claim 1, wherein the numerical attribute of the labeled data object comprises a recovery value that describes a financial or storage impact of a positive classification prediction for the labeled data object (Paragraphs 54 & 55 – teaches recovery score may be numerical value between any range of numbers, percentage, classification or other measurements, the score presents the likelihood the merchant/issuer will recover the transaction value or a portion thereof & the fraud risk score represents a likelihood that the requested transaction is fraudulent and may itself be expressed as a classification such as low risk/high risk/neutral risk, Juneja).
Claims 11-14 and 17-20 are similar to claims 1-4, 8-10 hence similarly rejected.
Claims 5-7 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Chowdhury et al. (US 11,138,523) in view of Gunda et al. (US 2023/0028278), Sachdev et al. (US 11,386,365) and Juneja et al. (US 2021/0304206) further in view of Steiner et al. (US 10,438,113)
All the limitations of claim 1 are taught above.
5. The combination of Chowdhury, Gunda, Sachdev and Juneja do not explicitly teach, wherein generating the target object embedding comprises:
receiving a cohort embedding for the unlabeled data object that corresponds to a cohort identifier associated with a cohort subset of the set of data objects that comprises the unlabeled data object; and
generating the target object embedding for the unlabeled data object based on a set of object features that comprises the cohort embedding.
However, Steiner teaches,
receiving a cohort embedding for the unlabeled data object that corresponds to a cohort identifier associated with a cohort subset of the set of data objects that comprises the unlabeled data object (Col 4:lines 36-45 – teaches generates, from the input data 100, a respective operation embedding for each of the plurality of operations. Generally, for each of the plurality of operations, the system 100 generates a respective property embedding for each of a plurality of properties of the operation and combines the property embeddings to generate the operation embedding. The properties include one or more of an operation type of the operation, a size of outputs generated by the operation, or adjacency information for the operation – disclose creating embeddings based on identifiers/properties of entities (operations)); and
generating the target object embedding for the unlabeled data object based on a set of object features that comprises the cohort embedding (Claim 1 – teaches assigning each of the operation groups to a respective device from the plurality of hardware devices, comprising: generating, for each of the operation groups, a respective group embedding by combining the operation embeddings for the operations in the operation group, and processing the respective group embeddings for the operation groups using a placer neural network having a plurality of placer parameters – discloses that embeddings can be combined into higher level embeddings (operation embeddings into group embeddings (i.e. cohort embeddings)).
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to allow Chowdhury to be combined with Steiner because Chowdhury, Sachdev and Juneja are in the same field of endeavor of machine learning systems that generate, process and combine embeddings for decision making tasks and by combining the prior arts would allow the cohort-level features of Steiner to enrich object embeddings in Chowdhury, which are then ranked or prioritized with classification/recovery values (Juneja).
Claim 15 is similar to claim 5 hence rejected similarly.
6. The combination of Chowdhury, Juneja, Sachdev and Steiner teach, The computer-implemented method of claim 5, wherein the set of object features further comprise at least one code embedding corresponding to a defined code within a code ontology, a code attribute associated with the defined code, or a count of defined codes from the code ontology that are detected within the unlabeled data object (Col 4: lines 36-43 - teaches the system 100 first generates, from the input data 100, a respective operation embedding for each of the plurality of operations. Generally, for each of the plurality of operations, the system 100 generates a respective property embedding for each of a plurality of properties of the operation and combines the property embeddings to generate the operation embedding – discloses creating embeddings for properties (code/attributes of operations), Steiner).
7. The combination of Chowdhury, Juneja, Sachdev and Steiner teach, The computer-implemented method of claim 5, wherein generating the target object embedding for the unlabeled data object comprises concatenating the set of object features based on a defined feature sequence (Col 4: lines 40-43 – teaches - the system 100 generates a respective property embedding for each of a plurality of properties of the operation and combines the property embeddings to generate the operation embedding – discloses combining property embeddings (features) into a single embedding, which is done by concatenating features in a defined sequence before feeding them into a model, Steiner) and encoding the concatenated set of object features (Claim 1 - generating, for each of the operation groups, a respective group embedding by combining the operation embeddings for the operations in the operation group, and processing the respective group embeddings for the operation groups using a placer neural network having a plurality of placer parameters – discloses after concatenation/combination encoding the embeddings through a neural network (placer NN), which corresponds to encoding the concatenated features into a target embedding, Steiner).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Dirac et al. (US 2015/0379428) – teaches quantile binning of numerical attributes, by selecting observation records for quantile binning transformations, with multiple bin counts applied to a particular variable (Abstract).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMRESH SINGH whose telephone number is (571)270-3560. 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, Ann J. Lo can be reached at (571) 272-9767. 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.
/AMRESH SINGH/Primary Examiner, Art Unit 2159
Prosecution Insights