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 .
Claims 1 – 20 are pending.
Response to Arguments
Applicant presents the following arguments in the 18 June 2026 amendment:
Accordingly, because the claims are directed to a specific technological improvement in search system functionality, integrate any alleged abstract idea into a practical application, and recite a non-conventional combination of elements that provides an inventive concept, the 101 rejection should be withdrawn.
First, Kazi does not describe conditionally executing a neural network model in response to determining that a query is a "null and low" query, as required by claim 1. Claim 1 involves determining that a query yields an insufficient number of results and, in response to that determination, triggering a reformulation workflow involving a modified neural model. Kazi, in contrast, merely describes that a user may be "not satisfied with the search result," and that alternative queries may be presented or generated. This generalized dissatisfaction signal is not equivalent to a system-level determination based on a quantitative threshold of retrieved results, nor does Kazi disclose gating execution of a particular neural architecture based on such a determination. Instead, Kazi's suggestion generation is part of a broader, always-available query suggestion framework, not a conditional recovery mechanism specifically triggered by null or low- result queries. The Office Action's mapping thus improperly equates fundamentally different triggers and system behaviors.
Kazi also does not describe the claimed "injecting a plurality of decoders into the neural network model" to generate reformulated queries. Claim 1 involves a specific architectural modification that involves augmenting a sequence-to-sequence model with multiple decoders that are jointly operated to produce distinct reformulations. Kazi, by contrast, describes a conventional sequence-to-sequence model that may include one or more layers of decoding units (e.g., LSTM units), but these are components of a single decoder pipeline that generates a single sequence output through sequential token prediction. The Office Action's reliance on Kazi's disclosure of "single and/or two layer decoders" conflates internal layering within a single decoder with the claimed plurality of independent decoders operating in parallel to produce multiple outputs. There is no description in Kazi of duplicating decoder structures to generate multiple reformulations concurrently, nor of coordinating such decoders in the manner required by the claims.
Next, Kazi fails to describe the claimed "diversity inducing optimization function that penalizes similarity between outputs of the plurality of decoders." The present specification explains that the model is explicitly trained and executed to produce reformulations that are not merely accurate, but intentionally diverse, by introducing an optimization term that penalizes similarity between outputs and enforces divergence among reformulations. This diversity-inducing mechanism supports the ability to capture multiple plausible interpretations of an ambiguous query. Kazi contains no such description. While Kazi may generate multiple query suggestions and may reference "different interpretations," any variation arises from probabilistic decoding or ranking mechanisms applied to a single model output, not from an explicit optimization objective that enforces diversity across multiple decoder outputs. The Office Action's assertion that suggested queries based on different interpretations is a diversity is conclusory and unsupported, as it fails to identify any disclosure in Kazi of a mechanism that penalizes similarity or otherwise enforces diversity as an optimization constraint. The claimed diversity-inducing optimization is therefore not present in Kazi.
Examiner presents the following responses to Applicant’s arguments:
With respect to applicant’s argument A, Applicant’s arguments have been fully considered and are persuasive in view of the amended claim language. The 35 USC 101 rejection of claims 1 – 20 has been withdrawn.
With respect to applicant’s argument B, Applicant's arguments have been fully considered but they are not persuasive. Kazi discloses in [0002] that a user not being satisfied with the results “may indicate that no results were identified or may include one or more results, none of which the user deems sufficiently relevant.” Further the claim states that a null or low query is correlated with the number of search results, but does not disclose the exact determination of what null or low comprises with regards to the number of search results, and further does not preclude the number of results being a number of acceptable or satisfactory results. Further, while the determination of the null or low query may be made at the search engine, as amended, however, this also does not explicitly limit to an automated function that does not rely on user feedback in the making of the determination. Further clarification to the claim language may be sufficient to limit the language to such an interpretation of a system-level automated determination without user input, however, the current language is not limited so.
With respect to applicant’s argument C, Applicant's arguments have been fully considered but they are not persuasive. Kazi discloses the generation of multiple reformulated queries from the single or dual layer of decoders in a sequence-to-sequence model with the neural network, each comprising a stack of multiple of LTSM units operating to generate outputs for generation of reformulated queries, see Kazi: [0064] – [0075], [0103] – [0110]. Additionally, the claim language is silent as to parallel and concurrent operations of the decoders. Clarification within the claim language to disclose the process for injection of decoders into the model as well as the relationship of each of the plurality of decoders to the plurality of reformulated queries, such as each separate decoder generating a separate reformulated query operating concurrently, over the disclosure of Kazi’s multi-decoder layers may overcome the current interpretation and mapping to Kazi.
With respect to applicant’s argument D, see rejection below in view of Kazi modified by Bodigutla below. Bodigutla discloses the use of incorporating a diversity measurement into the reward function for re-formulated search option options reduces the likelihood of the system choosing a repetitive utterance as a re-formulated search candidate, see Bodigutla: Para. 0075, 0122 – 0124, 0157.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 1 – 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication No. 2022/0172040 issued to Kazi et al (hereinafter referred to as Kazi) in view of US 20220100756 A1 issued to Bodigutla et al (hereinafter referred to as Bodigutla).
As to claim 1, Kazi discloses a method comprising:
receiving, by a search engine executing on at least one of one or more server devices of a search system, a query input from a user device (input search query, see Kazi: Para. 0002, 0014 – 0015, 0103 – 0110, and searcher may search the document database which comprises multiple databases, see Kazi: Para. 0032);
retrieving, from one or more databases of the search system, a number of items in response to the query input (receiving initial search results, see Kazi: Para. 0002, 0014 – 0015, 0103 - 0110);
determining, by the search engine, whether the query input is a null and low query based on the number of retrieved items (no results or user is not satisfied with the results, see Kazi: Para. 0002, 0014 – 0015, 0029, 0103 - 0110);
in response to the query input being the null and low query:
executing a neural network model by injecting a plurality of decoders into the neural network model and leveraging a diversity inducing optimization function to generate a plurality of diverse reformulated queries (based on the feedback that the user is not satisfied, generating reformulated query suggestions by way of machine learning, see Kazi: Para. 0015 – 0019 and 0059 - 0060, and generating the suggested queries using the query suggestion generator based on output from a neural network and sequence-to-sequence modeling, utilizing single and/or two layer decoders each comprising multiple LSTM units, each LSTM unit predicting an output, see Kazi: Para. 0064 – 0075, 0103 - 0110, and the query suggestions generated by use of the trained model, using optimization algorithms, to take into account different interpretations of different types of data, see Kazi: Para. 0019, 0049, 0083 – 0085, 0101 – 0110, suggested queries based on different interpretations is a diversity);
retrieving, from the one or more databases of the search system, a plurality of query results in response to the plurality of diverse reformulated queries (receiving and providing results for one or more suggested queries, including anticipatory search results before selection of a suggested query, see Kazi: Para. 0037, 0101 – 0110, and searcher may search the document database which comprises multiple databases, see Kazi: Para. 0032); and
providing, from the one or more service devices of the search system to the user device, the plurality of query results (providing results for one or more suggested queries, including anticipatory search results before selection of a suggested query, see Kazi: Para. 0037, 0101 – 0110).
However, Kazi does not explicitly disclose executing a neural network model by injecting a plurality of decoders into the neural network model and leveraging a diversity inducing optimization function that penalizes similarity between outputs of the plurality of decoders to generate a plurality of diverse reformulated queries.
Bodigutla teaches executing a neural network model by injecting a plurality of decoders into the neural network model and leveraging a diversity inducing optimization function that penalizes similarity between outputs of the plurality of decoders to generate a plurality of diverse reformulated queries (query reformulation options including prioritizing diversity between queries, including incorporating a diversity measurement into the reward function for re-formulated search option options reduces the likelihood of the system choosing a repetitive utterance as a re-formulated search candidate, see Bodigutla: Para. 0075, 0122 – 0124, 0157).
Bodigutla and Kazi are analogous due to their disclosure of sequence-to-sequence decoders in a neural network for generation of reformulated queries.
Therefore, it would have been obvious to one of ordinary skill in the art to modify Kazi’s use of generating reformulated queries using decoders placed into sequence-to-sequence models after receiving unsatisfactory number of results with Bodigutla’s use of diversifying queries output by decoders with sequence-to-sequence models of a neural network in order to improve a search engine's ability to deliver highly relevant search results in as few user interface-driven iterations as possible, see Bodigutla [0022].
As to claim 2, Kazi discloses the method of claim 1, wherein the plurality of query results is provided to the user device in a user interface without providing the plurality of diverse reformulated queries to the user device (displaying results is a designated query result portion of the interface, separate from the query suggestion portion of the user interface, see Kazi: Para. 0028 – 0029, and providing anticipatory search results, in a live manner, for the query suggestions as a user enters an incomplete original query, the query suggestions only being displayed after user selection of an anticipatory result or indicated interest in an anticipatory result, see Kazi: Para. 0037, the query suggestions results are displayed live before the query suggestions are displayed).
As to claim 3, Kazi discloses the method of claim 1, wherein the plurality of query results and the plurality of diverse reformulated queries are provided to the user device in a user interface (providing suggested queries and results (including the anticipatory results) to the user device through the user interface, see Kazi: Para. 0029, 0037, 0101).
As to claim 4, Kazi discloses the method of claim 3, further comprising separating each of the plurality of diverse reformulated queries and the corresponding plurality of query results within the user interface (suggested queries are displayed in a query suggestion portion of the user interface, and the received anticipatory results are displayed when a user selects the suggest query, see Kazi: Para. 0029 – 0030, 0037, 0101 and Fig. 2).
As to claim 5, Kazi discloses the method of claim 1, further comprising: training a sequence-to-sequence model utilizing historical user data (using the original query, user feedback queries and suggested queries, stored in a query history log, to further train a ML model, including a sequence-to-sequence model, see Kazi: Para. 0014 – 0020, 0034 – 0035, 0042 – 0044, 0049, 0101 – 0110); and
injecting the plurality of decoders in the sequence-to-sequence model (generating the suggested queries using the query suggestion generator based on output from a neural network and sequence-to-sequence modeling, utilizing single and/or two layer decoders each comprising multiple LSTM units, each LSTM unit predicting an output, see Kazi: Para. 0064 – 0075, 0101 - 0110).
As to claim 6, Kazi discloses the method of claim 5, wherein the historical user data comprises a search query and two query reformulations (using the original query, user feedback queries and suggested queries, stored in a query history log, to further train a ML model, including a sequence-to-sequence model, see Kazi: Para. 0014 – 0020, 0034 – 0035, 0042 – 0044, 0049, 0101 – 0110).
As to claim 7, Kazi discloses the method of claim 6, wherein each of the two query reformulations comprise one or more of dropped tokens, replaced tokens, or added tokens corresponding to the search query (query suggestion generator adds query tokens, different tokens, etc. to the embeddings for generation of suggested queries, see Kazi: Para. 0064 – 0065, 0106 – 0108).
Claims 8 and 15 are rejected using similar rationale to the rejection of claim 1 above.
Claims 9 and 16 are rejected using similar rationale to the rejection of claim 2 above.
Claims 10 and 17 are rejected using similar rationale to the rejection of claim 3 above.
Claims 11 and 18 are rejected using similar rationale to the rejection of claim 4 above.
Claims 12 and 19 are rejected using similar rationale to the rejection of claim 5 above.
Claims 13 and 14 are rejected using similar rationale to the rejection of claims 6 and 7 above, respectively.
Claim 20 is rejected using similar rationale to the rejection of claims 6 and 7 above.
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MARK E HERSHLEY whose telephone number is (571)270-7774. The examiner can normally be reached M-F: 9am-6pm.
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, Amy Ng can be reached at (571) 270-1698. 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.
/MARK E HERSHLEY/Primary Examiner, Art Unit 2164
Prosecution Insights