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 .
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office Action has been withdrawn pursuant to 37 CFR 1.114.
Detailed Action
This Non-Final Office Action is responsive to Applicants’ RCE submission and IDS dated 3/12/26, and is a result of prosecution being reopened on 3/31/26 as a result of the Examiner’s consideration of that same IDS.
Claims 1, 3-8, 10-15, and 17-20 remain pending, of which claims 1, 8, and 15 are independent.
Claim Rejections - 35 USC § 103
5. 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.
6. 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.
7. The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
8. Claims 1, 3, 8, 10, 15, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Non-Patent Literature “Knowledge Questions from Knowledge Graph” (Seyler) in view of CN 104471568 A (Zhou, as recently cited in Applicants’ IDS dated 3/12/26) and further in view of Non-Patent Literature “Knowledge graph fusion for smart systems: A survey” (Nguyen).
Regarding claim 1, SEYLER teaches a computer-implemented method for training a machine learning system (automatic generation of questions using knowledge graphs, e.g. as discussed in the Abstract on page 11 and the Conclusion on page 18, and elaborated on in sections 1-2 on pages 11-12, constitutes “a computer-implemented method” as recited, and the aforementioned aspects are integrated into a larger framework for training a “model” as discussed per section 4.2 on page 14 and the Conclusion on page 18 for example), the computer-implemented method comprising:
obtaining a current data structure from a ... knowledge graph ... the current data structure ... including a current head element, a current relationship element, and a current tail element (“knowledge graph”/”KG” as discussed in section 2 on page 12, the KG including instances of triples, where a triple constitutes the linking of entities by a predicate, e.g. such as subject-predicate-object, which would feasibly relate a subject and object together as a function of the predicate, which the Examiner equates with the recited “current data structure” obtainable from a recited “knowledge graph”, and where subject/predicate/object as taught reads on “head”/”relationship”/”tail” elements as recited);
obtaining a sentence corresponding to the current data structure (query generation, per section 3 beginning on page 12, involves generating a query for a unique answer stored in and obtained from the KG’s “question-answer” corpus (section 2, subheading “Generality”, on page 12), where the corpus’s Q-A constitutes a “fact” (section 2, subheading “Knowledge Graph” first paragraph, page 12), which the Examiner equates with the recited “sentence”);
generating a question by removing the current tail element from the sentence and generating a correct answer to the question, the correct answer including the current tail element (query generation, per section 3 as continued on page 13, discussing drawing triple patterns from the KG where the subject or object in the drawn pattern matches the particular answer for the query to be generated, and where the drawn patterns could feasibly necessarily include the pattern that already includes the initial question-answer item discussed above per the limitation “obtaining ...”);
extracting a pool of data structures from the ... knowledge graph … to … create a pool of distractor candidates (distractor generation, per section 6.1 on page 15, involves the query of section 3, as discussed above, which involves drawing triple patterns from a KG (e.g. equivalent to the recited “extracting ... data structures” step addressed here), and where the entities from the drawn triple patterns constitute distractors);
selecting a set of distractors from the pool of distractor candidates (section 6.1 discussing restrictions such as type and distance, which are used to reduce a first set of distractors to a smaller set of distractors which are more salient to the query);
creating a query task that includes the question and a set of response options, the set of response options including the correct answer and the set of distractors and creating a training set that includes at least the query task and training the machine learning system with the training set, wherein the machine learning system is configured to receive the query task and respond to the question with a predicted answer that is selected from among the set of response options (staying on page 15, section 6.1’s distractor generation results in the derivation of additional queries that make use of the distractors to arrive at Q’ (as opposed to the original query Q)).
Regarding the generation of a distractor pool as discussed above, the Examiner does not believe Seyler teaches the further limitation of extracting a pool of data structures from the ... knowledge graph … to … create a pool of distractor candidates specifically based on a set of distractor criteria which the Examiner believes is further clarified where each extracted data structure having a head element with no common keywords with the current head element and a relationship element that is the current relationship element ... and extracting tail elements from the pool of data structures to create a pool of distractor candidates. Rather, the Examiner relies upon ZHOU to teach what Seyler otherwise lacks, see e.g., Zhou’s page 5 discussing “the fourth stage” of “answer extraction/sorting”, in what appears to be that page’s 5th full paragraph, where specifically candidate answers have been extracted having answers from the returned search results, where the answers are akin to extracted tail elements from a pool that are used as distractor candidates, and where the relationship element for these search results appear to be the same, e.g. some version of being a composer.
Regarding the knowledge graph as discussed above in relation to the claim and also Seyler, the Examiner notes that Applicants’ claim clarifies that the knowledge graph is a global knowledge graph that includes a combination of various knowledge graphs and that the obtained current data structure is one being associated with one particular knowledge graph of the global knowledge graph, which Seyler does not teach but Zhou as discussed just above does: see, e.g., Zhou’s page 5 discussing “the third stage … evidence collection” which is a consolidation of search results from different queries, i.e., the combining of information from various knowledge graphs, such that the consolidated product could be understood to be akin to the recited global knowledge graph.
Based on the reasoning provided above, the Examiner believes that Seyler modified in view of Zhou may teach the amended limitations as provided by way of Applicants’ recent RCE. The reasoning is based on the Examiner’s understanding of Zhou’s consolidation of search results, where the results as consolidated would be understood to correspond to different knowledge graphs for example. However, to the extent that this is not reasonably extensible to teach the amended limitations provided by way of the recent RCE, the Examiner further relies upon NGUYEN to teach what Seyler etc. may otherwise lack, see e.g. Nguyen’s pages 56-57 discussing an approach to encompass knowledge representation from explicitly disparate data sources that is inclusive of “correlating and combining different big data sources and existing knowledge graphs to construct an updated knowledge graph”, which Nguyen terms “knowledge graph fusion.” Further elaboration is provided in section 2.2’s “Approach 2 (A2).” From a reading of Nguyen, the notion of combining different knowledge graphs together into a unified knowledge graph is known in the state of the art, and in the Examiner’s view reads on Applicants’ recitation of a global knowledge graph that is capable of having embeddings that would originate from different component knowledge graphs that have been combined or integrated to generate that global knowledge graph. Moreover, the idea that these embeddings are foundational to model generation, training, testing, etc. is well-understood and practiced in the state of the art.
Similar to Seyler and arguably Zhou, Nguyen contemplates knowledge graph driven processes that permit the accrual of knowledge in useful formats to help drive machine learning approaches and techniques, e.g. to solve problems, provide recommendations, etc. Hence, the aforementioned references are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to broaden Seyler’s modified framework with Nguyen’s approach to build a more encompassing knowledge representation, with a reasonable expectation of success, such that a broader sweep of knowledge could feasibly be used to benefit improved Q & A generation, particularly by helping the model and its training with exposure to more knowledge.
Regarding claim 3, Seyler in view of Zhou and further in view of Nguyen teach the computer-implemented method of claim 1, as discussed above. The aforementioned references teach the additional limitation wherein each extracted data structure has a tail element that is not found in another data structure of the global knowledge graph in which the another data structure includes the current head element and the current relationship element (Seyler’s section 3 on page 13, first column, first full paragraph discussing the use of a unique answer for the generated query, which the Examiner equates with a tail element being unique as recited). The motivation for combining the references is as discussed above in relation to claim 1.
Regarding claim 8, the claim includes the same or similar limitations as claim 1 discussed above, and is therefore rejected under the same rationale. The claim additionally recites a data processing system comprising at least one non-transitory computer readable medium including a neuro-symbolic framework and a processing system, which are further taught per Seyler’s model-based training aspect and specifically its knowledge graphs, and where any framework that functions as described in Seyler much necessarily feature fundamental computer elements such as processors and memory as widely understood and practiced in the state of the art.
Regarding claim 10, the claim includes the same or similar limitations as claim 3 discussed above, and is therefore rejected under the same rationale.
Regarding claim 15, the claim includes the same or similar limitations as claim 1 discussed above, and is therefore rejected under the same rationale. The claim additionally recites a computer product comprising at least one non-transitory computer readable storage device, which are further taught per Seyler’s model-based training aspect and specifically its knowledge graphs, and where any framework that functions as described in Seyler much necessarily feature fundamental computer elements such as processors and memory as widely understood and practiced in the state of the art.
Regarding claim 17, the claim includes the same or similar limitations as claim 3 discussed above, and is therefore rejected under the same rationale.
9. Claims 4, 11, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Seyler in view of Zhou and further in view of Nguyen and further yet in view of WO 2020/072194 A2 (Goldman).
Regarding claim 4, Seyler in view of Zhou and further in view of Nguyen teach the computer-implemented method of claim 1, as discussed above. The aforementioned references teach the additional limitation wherein the step of selecting the set of distractors from the pool of distractor candidates and creating the set of distractors to include a subset of the chosen distractor candidates, e.g. for the reasons discussed above in relation to claim 1, but not choosing distractor candidates via a random selection process. Rather, the Examiner relies upon GOLDMAN to teach what Seyler etc. otherwise lack, see e.g. Goldman’s [0036] and [0043].
Like Seyler, Goldman is directed to question and answer generation, and specifically through the use of determinable distractors. Hence, the aforementioned references are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Goldman’s distractor selection aspect into Seyler’s framework and similar notion of distractor determination/usage, with a reasonable expectation of success, such that a normal and widely understood selection mechanism known in the computing arts, e.g. randomization, could be applied to specifically implement a feature that Seyler otherwise contemplates. In other words, it would be obvious to try.
Regarding claim 11, the claim includes the same or similar limitations as claim 4 discussed above, and is therefore rejected under the same rationale.
Regarding claim 18, the claim includes the same or similar limitations as claim 4 discussed above, and is therefore rejected under the same rationale.
10. Claims 5-6, 12-13, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Seyler in view of Zhou and further in view of Nguyen and further yet in view of US 2018/0061256 (“Elchik”).
Regarding claim 5, Seyler in view of Zhou and further in view of Nguyen teach the computer-implemented method of claim 1, as discussed above. The aforementioned references teach the additional limitation wherein the step of selecting the set of distractors from the pool of distractor candidates and creating the set of distractors to include a subset of the chosen distractor candidates, e.g. for the reasons discussed above in relation to claim 1, but not choosing distractor candidates that have a greatest cosine similarity score with respect to the correct answer while satisfying at least one threshold. Rather, the Examiner relies upon Elchik to teach what Seyler etc. otherwise lack, see e.g. Elchik’s [0086] discussing the use of cosine similarity for purposes of generating both question and answer element distractors such as further clarified in at least [0089] and [0091] for example.
Like Seyler, Elchik is directed to question and answer generation, and specifically through the use of determinable distractors. Hence, the aforementioned references are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Elchik’s distractor selection aspect into Seyler’s framework and similar notion of distractor determination/usage, with a reasonable expectation of success, such that a normal and widely understood selection mechanism known in the computing arts, e.g. particularly among similarity metrics, could be applied to specifically implement a feature that Seyler otherwise contemplates. In other words, it would be obvious to try.
Regarding claim 6, Seyler in view of Zhou and further in view of Nguyen teach the computer-implemented method of claim 1, as discussed above. The aforementioned references teach the additional limitation wherein the step of selecting the set of distractors from the pool of distractor candidates and creating the set of distractors to include a subset of the chosen distractor candidates, e.g. for the reasons discussed above in relation to claim 1, but not choosing distractor candidates that have a greatest cosine similarity score with respect to the question while satisfying at least one threshold. Rather, the Examiner relies upon Elchik to teach what Seyler etc. otherwise lack, see e.g. Elchik’s [0086] discussing the use of cosine similarity for purposes of generating both question and answer element distractors such as further clarified in at least [0089] and [0091] for example.
Like Seyler, Elchik is directed to question and answer generation, and specifically through the use of determinable distractors. Hence, the aforementioned references are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Elchik’s distractor selection aspect into Seyler’s framework and similar notion of distractor determination/usage, with a reasonable expectation of success, such that a normal and widely understood selection mechanism known in the computing arts, e.g. particularly among similarity metrics, could be applied to specifically implement a feature that Seyler otherwise contemplates. In other words, it would be obvious to try.
Regarding claim 12, the claim includes the same or similar limitations as claim 5 discussed above, and is therefore rejected under the same rationale.
Regarding claim 13, the claim includes the same or similar limitations as claim 6 discussed above, and is therefore rejected under the same rationale.
Regarding claim 19, the claim includes the same or similar limitations as claim 5 discussed above, and is therefore rejected under the same rationale.
Regarding claim 20, the claim includes the same or similar limitations as claim 6 discussed above, and is therefore rejected under the same rationale.
11. Claims 7 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Seyler in view of Zhou and further in view of Nguyen and further yet in view of Non-Patent Literature “Self-Supervised Knowledge Triplet Learning for Zero-Shot Question Answering” (“Banerjee”).
Regarding claim 7, Seyler in view of Zhou and further in view of Nguyen teach the computer-implemented method of claim 1, as discussed above. The aforementioned references teach the additional limitations for obtaining a task dataset having other tasks that are distinct from the query task (e.g., each query/question per Seyler and Zhou for example could be a distinct task, in which case training would be understood to be generalized to many queries/questions, and it would not make any sense to have a model that is only answerable to one specific query/question) … wherein the machine learning system is trained with the training set during a pre-training phase of the machine learning system (the algorithms mentioned in Zhou specifically which are understood to perform the ranking and sorting of distractor candidates are specifically said to be trained, and they are in support of the later machine learning task of training the question-answer processing to arrive at the correct answer, in which case the training of those elements earlier in the machine learning pipeline/process could be understood to be a pre-training training phase, relatively speaking) but do not teach performing a zero-shot evaluation of the machine learning system based on the task dataset. Rather, the Examiner relies upon Banerjee, see e.g., Abstract mentioning zero-shot evaluation in a comparable question-answer framework.
Like Seyler and Zhou, Banerjee relates to a similar question-answer framework, and is therefore similarly directed and hence analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Banerjee’s efficiencies to improve Seyler’s modified framework to function in a zero-shot evaluation capacity as recited.
Regarding claim 14, the claim includes the same or similar limitations as claim 7 discussed above, and is therefore rejected under the same rationale.
Conclusion
12. Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHOURJO DASGUPTA whose telephone number is (571)272-7207. The examiner can normally be reached M-F 8am-5pm CST.
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, Tamara Kyle can be reached at 571 272 4241. 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.
/SHOURJO DASGUPTA/Primary Examiner, Art Unit 2144
Prosecution Insights