DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This action is responsive to communications: Amendment filed on 1/2/2026.
Claims 1-20 are pending. Claims 1, 8, and 14 are independent.
The previous rejection of claims 1-20 under 35 USC § 103 have been maintained in view of the amendment.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-7 is/are rejected under 35 U.S.C. 103 as being unpatentable over Fauceglia et al. (US2018/0137404) in view of Lin et al. (“Progressive Joint Framework for Chinese Question Entity Discovery and Linking With Question Representations”) and Byrnes et al. (US2020/0193286).
In regards to claim 1, Fauceglia et al. substantially discloses a computer system comprising:
a processor operatively coupled to memory (Fauceglia et al. para[0010]);
an artificial intelligence (AI) platform, operatively coupled to the processor, comprising:
a machine learning (ML) manager, operatively coupled to the evaluator, configured to leverage an artificial neural network (ANN) and a corresponding ML algorithm to learn the connective weights (Fauceglia et al. para[0050], the same weight matrices for all time steps to be learned during training);
the ML manager configured to selectively update the connective weights associated with the logically connected rules (Fauceglia et al. para[0052], optimize parameters of the weight vectors);
generate a learned model with learned thresholds and the learned connective weights for the logically connected rules (Fauceglia et al. para[0037], weights are computed for feature vectors)
Fauceglia et al. does not explicitly disclose a feature manager to generate a set of features for one or more entity- mention pairs in an annotated dataset.
However Lin et al. substantially discloses a feature manager to generate a set of features for one or more entity- mention pairs in an annotated dataset (Lin et al. pg146283 section II.A para3, they build character-level, word-level or sequential features from annotated datasets and leverage machine learning algorithm to learn mention patterns).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the entity linking framework of Lin et al. in order to improve the use of annotated datasets (Lin et al. pg146283 section I para5).
Fauceglia et al. does not explicitly disclose an evaluator configured to evaluate the generated set of features of the one or more entity-mention pairs against an entity linking (EL) LNN rule template, the template having one or more logically connected rules and corresponding connective weights organized in a hierarchical structure.
However Byrnes et al. discloses an evaluator configured to evaluate the generated set of features of the one or more entity-mention pairs against an entity linking (EL) LNN rule template, the template having one or more logically connected rules and corresponding connective weights organized in a hierarchical structure (Byrnes et al. para[0041], These cost functions can be used to provide a confidence score indicating how well the current interpretation satisfies the rules and data);
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the adaptive network of Byrnes et al. in order to generate a machine learning architecture from a logical specification (Byrnes et al. para[0005]).
In regards to claim 2, Fauceglia et al. as modified by Lin et al. and Byrnes et al. discloses the system of claim 1, wherein the evaluation further comprises the evaluator to re- formulate an entity linking algorithm composed of a disjunctive set of rules into an LNN representation for learning, wherein the entity linking is a restricted form of first order logic rules comprising a set of Boolean predicates connected by logical operators (Lin et al. pg146286 section IV.A para5, the semantic node logic defines the logical combination relation of nodes and includes five types of logic relations: “entailment,” “and,” “or,” “not” and “list”).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the entity linking framework of Lin et al. in order to improve the use of annotated datasets (Lin et al. pg146283 section I para5).
In regards to claim 3, Fauceglia et al. as modified by Lin et al. and Byrnes et al. discloses the system of claim 2, wherein the entity-mention pair evaluation further comprises the evaluator to compute one or more features for a subset of labeled entity-mention pairs, wherein each of the features has a corresponding similarity predicate (Fauceglia et al. para[0044], The method groups the resulting global features into a single feature vector to compute the global similarity).
In regards to claim 4, Fauceglia et al. as modified by Lin et al. and Byrnes et al. discloses the system of claim 3, further comprising the ML manager to leverage the ANN and the ML algorithm to learn an appropriate threshold for each of the computed one or more features as related to the corresponding similarity predicate (Fauceglia et al. para[0032], the global component that runs recurrent neural networks on the entity mentions to generate the global features ϕ.sub.global at 195. Finally, at step 197, the method computes a final ranking score (threshold) as the sum of the scores obtained at steps 190 and 195).
In regards to claim 5, Fauceglia et al. as modified by Lin et al. and Byrnes et al. discloses the system of claim 4, further comprising the evaluator to filter the computed one or more features based on their corresponding learned threshold, and selectively incorporate the computed one or more features into the LNN rule template responsive to the filtering, the selective incorporation including removal of a feature or assignment of a non-zero score to the feature (Lin et al. pg166285 section III para2, the extracted mentions can be regarded as one feature that can be incorporated into another statistical learning ED method).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the entity linking framework of Lin et al. in order to improve the use of annotated datasets (Lin et al. pg146283 section I para5).
In regards to claim 6, Fauceglia et al. as modified by Lin et al. and Byrnes et al. discloses the system of claim 2, further comprising a rule manager, operatively coupled to the evaluator, configured to:
learn one or more of the logically connected rules (Byrnes et al. para[0029], This inputted knowledge is used to create that act as constraints on the subsequent learning that occurs in the created machine learning model training on data.);
dynamically generate a template for the hierarchical structure (Byrnes et al. para[0032], turn each statement into its own tree structure of nodes and then layers of nodes for each predicate and/or function in that statement);
learn a logical rule based on the dynamically generated template (Byrnes et al. para[0042], The artificial intelligence engine uses both the set of rules derived from the knowledge representations and reasoning and then adaptions to the understanding of those rules derived from training data to allow for fewer training cycles);
evaluate a selected rule on a labeled dataset (Byrnes et al. para[0046], adjust the understanding of each constant/entity, function and predicate in a set of rules, based on the data being analyzed); and
selectively assign the selected rule to a corresponding node in the hierarchical structure based on the evaluation (Byrnes et al. para[0081], constructs full network from formula parse trees).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the adaptive network of Byrnes et al. in order to generate a machine learning architecture from a logical specification (Byrnes et al. para[0005]).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the adaptive network of Byrnes et al. in order to generate a machine learning architecture from a logical specification (Byrnes et al. para[0005]).
In regards to claim 7, Fauceglia et al. as modified by Lin et al. and Byrnes et al. discloses the system of claim 6, wherein the template is a binary tree and the corresponding node is an internal node, and further comprising the rule manager to selectively assign a conjunctive or disjunctive LNN operator to the internal node (Byrnes et al. fig. 4 para[0075], The parser module may be configured to know from the syntax what category these different parts of the statement go into).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the adaptive network of Byrnes et al. in order to generate a machine learning architecture from a logical specification (Byrnes et al. para[0005]).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the adaptive network of Byrnes et al. in order to generate a machine learning architecture from a logical specification (Byrnes et al. para[0005]).
Claim(s) 8-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Fauceglia et al. in view of Lin et al., Byrnes et al. and Nickel et al. (“A review of Relational Machine Learning for Knowledge Graphs”).
In regards to claim 8, Fauceglia et al. substantially discloses a computer program product for disambiguating mentions in text, the computer program product comprising one or more non-transitory computer readable storage mediums having program code embodied therewith, the program code comprising programming instructions for:
Leveraging an artificial neural network (ANN) and a corresponding machine learning (ML) algorithm to learn the connective weights (Fauceglia et al. para[0050], the same weight matrices for all time steps to be learned during training).
Fauceglia et al. does not explicitly disclose generating features for one or more entity-mention pairs in an annotated dataset.
However Lin et al. discloses generating features for one or more entity-mention pairs in an annotated dataset (Lin et al. pg146283 section II.A para3, they build character-level, word-level or sequential features from annotated datasets and leverage machine learning algorithm to learn mention patterns).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the entity linking framework of Lin et al. in order to improve the use of annotated datasets (Lin et al. pg146283 section I para5).
However Fauceglia et al. does not explicitly disclose evaluating the generated features of the one or more entity-mention pairs against an entity linking (EL) logical neural network (LNN) rule template, the template having one or more logically connected rules and corresponding connective weights organized in a hierarchical structure.
However Byrnes et al. substantially discloses Evaluating the generated features of the one or more entity-mention pairs against an entity linking (EL) logical neural network (LNN) rule template, the template having one or more logically connected rules and corresponding connective weights organized in a hierarchical structure (Byrnes et al. para[0041], These cost functions can be used to provide a confidence score indicating how well the current interpretation satisfies the rules and data).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the adaptive network of Byrnes et al. in order to generate a machine learning architecture from a logical specification (Byrnes et al. para[0005]).
However Fauceglia et al. does not explicitly disclose Generating features for one or more entity-mention pairs in an annotated dataset, wherein the features correspond to attributes that measure a degree of similarity between a textual mention and a candidate entity;
formulating the LNN rule template with EL rules;
Training the LNN formulated EL rules over a labeled dataset using the ANN and the corresponding ML algorithm to perform gradient descent; and
Generating a learned model with learned thresholds and the learned connective weights for the logically connected rules thereby disambiguating mentions in short text by linking them to entities in a logical neural network using interpretable rules, wherein the short text comprises a single sentence or question.
Nickel et al. substantially discloses generating features for one or more entity-mention pairs in an annotated dataset, wherein the features correspond to attributes that measure a degree of similarity between a textual mention and a candidate entity (Nickel et al. pg14 section II.E para2-3, uses features and links to determine the similarity between an object and an underlying entity);
formulating the LNN rule template with EL rules (Nickel et al. pg22 section V.B para1, extracts logical rules from EL rules (knowledge graph) pg29 section X.C para1, using weights to calculate a new relationship);
training the LNN formulated EL rules over a labeled dataset using the ANN and the corresponding ML algorithm to perform gradient descent (Nickel et al. pg18 section IV.A para6-7, uses gradient descent algorithm to train model over a small number of iterated updates); and
generating a learned model with learned thresholds and the learned connective weights for the logically connected rules thereby disambiguating mentions in short text by linking them to entities in a logical neural network using interpretable rules, wherein the short text comprises a single sentence or question (Nickel et al. pg29 XI para1, generates a learned model for interpreting and answering questions).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the entity linking method of Fauceglia et al. with the knowledge graphs of Nickel et al. in order to web of data readable by machines (Nickel et al. pg12 section II.A para1).
Claim 9 recites substantially similar limitations to claim 2, claims 10 recites substantially similar limitations to claim 3, claim 11 recites substantially similar limitations to claims 4-5, claim 12 recites substantially similar limitations to claim 6, and claim13 recites substantially similar limitations to claim 7. Thus claims 9-13 are rejected along the same rationale as claims 2-7.
Claim 14 recites substantially similar limitations to claim 8. Thus claim 14 is rejected along the same rationale as claim 8.
Claims 15-20 recite substantially similar limitations to claims 2-7. Thus claims 15-20 are rejected along the same rationale as claims 2-7.
Response to Arguments
Applicant’s arguments with respect to claims 1-20 have been considered but are moot because the arguments do not apply the current rejection.
Applicant argues on pg2 that Fauceglia et al. does not teach "generate a set of feature for one or more entity-mention pairs in an annotated dataset".
However Fauceglia et al. in view of Lin et al. and Byrnes et al. discloses generate a set of feature for one or more entity-mention pairs in an annotated dataset (Lin et al. pg146283 section II.A para3, build sequential features from annotated datasets to learn mention patterns, uses part of speech tags, word affixes, and context as features.)
Applicant argues on pg3 that Fauceglia et al. does not teach "a machine learning (ML) manager, operatively coupled to the evaluator, configured to leverage an artificial neural network (ANN) and a corresponding ML algorithm to learn the connective weights".
However Fauceglia et al. in view of Line et al. and Byrnes et al. discloses a machine learning (ML) manager, operatively coupled to the evaluator, configured to leverage an artificial neural network (ANN) and a corresponding ML algorithm to learn the connective weights (Fauceglia et al. para[0050], leverages Recurrent Neural Network to learn connective weights W, U, and V during training).
Applicant argues on pg6 that Fauceglia et al. does not teach "generate a learned model with learned thresholds and the learned connective weights for the logically connected rules"
However Fauceglia et al. in view of Line et al. and Byrnes et al. discloses generate a learned model with learned thresholds and the learned connective weights for the logically connected rules (Fauceglia et al. para[0037], computes weights for feature vectors that captures linguistic properties and statistics that have been discover for Entity Linking)
Applicant argues on pg7 that Fauceglia et al. does not teach "formulating the LNN rule template with EL rules"
However Fauceglia et al. as modified by Lin et al., Byrnes et al., and Nickel et al. discloses formulating the LNN rule template with EL rules (Nickel et al. pg22 section V.B para1, attempts to learn rules from relational data via inverse entailment, pg29 section X.C para1, calculates representation of new entities based on observed relationships)
Applicant argues on pg9 that Fauceglia et al. does not teach "training the LNN formulated EL rules over a labeled dataset using the ANN and the corresponding ML algorithm to perform gradient descent"
However Fauceglia et al. as modified by Lin et al., Byrnes et al., and Nickel et al. discloses training the LNN formulated EL rules over a labeled dataset using the ANN and the corresponding ML algorithm to perform gradient descent (Nickel et al. pg18 section IV.A para 6-7, Uses RESCAL-ALS algorithm to efficiently calculate parameters using gradient descent)
Applicant also argues that Fauceglia et al. does not teach "generating a learned model with learned thresholds and the learned connective weights for the logically connected rules thereby disambiguating mentions in text by linking them to entities in a logical neural network using interpretable rules".
However Fauceglia et al. as modified by Lin et al., Byrnes et al., and Nickel et al. discloses generating a learned model with learned thresholds and the learned connective weights for the logically connected rules thereby disambiguating mentions in text by linking them to entities in a logical neural network using interpretable rules (Nickel et al. pg29 section XI para1, uses statistical relational learning to generate very large knowledge graphs used for question answering, structured search, exploratory search, and digital assistants).
Applicant argues on pg13 that Fauceglia et al. does not teach "generating features for one or more entity-mention pairs in an annotated dataset, wherein the features correspond to attributes that measure a degree of similarity between a textual mention and a candidate entity".
However Fauceglia et al. as modified by Lin et al., Byrnes et al., and Nickel et al. discloses (Nickel et al pg14 section II.E para2-3, generate features to cluster entities and mentions based on similarity).
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS HASTY whose telephone number is (571)270-7775. The examiner can normally be reached Monday-Friday 8:30am-5:00pm.
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, Matt Ell can be reached at (571)270-3264. 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.
/N.H/Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141
Prosecution Insights