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
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Applicant’s amendments to the claims filed 5/26/26 suffice to obviate the rejection of claims 2-6, 10-14, 16-20 under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as discussed in the Non-Final action of 5/7/26.
Newly amended claims 1, 3, 4, 8, 9, 11, 12, 15, 17, 18, 21 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention. Claims 1, 9, 15 recite both “a natural language query,” and “an input natural language query,” but subsequently recite processing “contextual features from the natural language query,” then additionally recite that the system responds to “the input natural language query.” While parseable, the same or similar query(s) are referred to by two different names which renders the claim indefinite. Claims 1, 9, 15 additionally recite “a dataset of natural language phrases, which is associated with a certain category, and a training dataset configured for training a machine learning model.” The inclusion renders the resolved subject matter to modify ambiguous, is the dataset of phrases modified or the desired overlap? Claims 1, 9, 15 additionally recite “identifying a skill or an intent that are associated with the chatbot system and are usable by the chatbot system.” The identifying resolves a singular item but the claim uses a plural verb; this grows particularly indefinite with respect to claim 21 which recites the claimed chatbot operable to select the identified skill or intent; does claim 21 select a set or a singular item? Claims 3, 4, 8, 11, 12, 15, 18 do not remedy and are similarly rejected. Appropriate correction is required.
Claim Rejections - 35 USC § 101
Applicants amendments suffice to obviate the 35 U.S.C. 101 of Claims 1, 3, 4, 8, , 9, 11, 12, 15, 18, 21 because the claimed invention is considered to integrate any judicial exception into an improvement to a chatbot.
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.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied 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.
Claims 1, 3, 4, 8, , 9, 11, 12, 15, 18, 21 rejected under 35 U.S.C. 103 as being unpatentable over Goyal: Fast and Scalable Expansion of Natural Language Understanding Functionality for Intelligent Agents (copy provided by Examiner; copyright 2018; and hereinafter Go) further in view of Song: Improving Neural Named Entity Recognition with Gazetteers (copy provided by Examiner; copyright 2020; and hereinafter So) and further in view of Kang: 10303978 hereinafter Ka.
Regarding claim 1
Go teaches:
A computer-implemented method performed by at least one computer system, the computer-implemented method (Go: Abstract; Fig 1: a neural network architecture for managing operations of a chatbot) comprising:
a dataset of natural language phrases, which is associated with a certain category (Go: § 3, pp 2: system creates a dataset comprising a categorical list of slot values; “developers need to create gazetteers for their domain, which are lists of slot values,”; such as FoodItem-->chicken, etc.), and
a training dataset configured for training a machine learning model that utilizes the dataset of natural language phrases for processing queries for a chatbot system (Go: § 3, pp 2, 3; § 4.1, pp 3; Fig 1: system trains a conversational agent suing the categorical list of slot values; fine turns using additional features of the categorical list);
generating one or more training pairs from the one or more natural language phrases, each of the training pairs including a natural language query generated from a respective natural language phrase (Go: § 3, pp 2; § 4.1, pp 3, 4; § 4.2, pp 4; § 5.1, pp 5: system generates labelled training utterances by sampling categorical list values such that each comprises an intent/slot label);
retraining the machine learning model using the modified training dataset (Go: § 4.1, pp 3, 4; § 4.2, pp 4; Fig 1: such as retraining by fine tuning by replacement of data); and
processing, using the retrained machine learning model, an input natural language query received by the chatbot system (Go: § 4.1, pp 2, 3: such as by performing the disclosed method on a given input user utterance),
wherein the retrained machine learning model processes contextual features generated from the natural language query and expressional features generated from the dataset of natural language phrases (Go: § 4.1, pp 2, 3; Fig 1: such as using the model fine-tuned with the additional, replaced, etc. features by concatenating contextual features such as word/embedding type features with expressional features from the categorical list or sampled categorical list data)
to generate an output identifying a skill or an intent that are associated with the chatbot system and are usable by the chatbot system to respond to the input natural language query (Go: § 3, pp 2, 3; § 4.1, pp 3, 4: such as by output of an intent label with respect to a developer defined skill).
Go does not explicitly teach coverage values, particularly receiving an indication of a first desired coverage value corresponding to a desired overlap between a dataset of natural language phrases, which is associated with a certain category, and a secondary dataset wherein the system further operates for determining a second coverage value by determining a number of natural language phrases from the dataset of natural language phrases that are present in the training dataset and associated in the training dataset with a gold label category matching the certain category; such that in response to determining that the second coverage value is less than the first coverage value, modifying the training dataset to include one or more natural language phrases from the dataset of natural language phrases the modifying including: selecting, one of from the dataset of natural language phrases, the one or more natural language phrases that are not present in the training dataset; selecting, from the dataset of natural language phrases, the one or more natural language phrases that are not present in and the training dataset; and the generating training pairs including the gold label category matching the certain category, domain, etc.
In a related field of endeavor So teaches a system for using gazetteers such as the categorical list of Go to improve entity recognition (So: Abstract) comprising receiving an indication of a first coverage value corresponding to a desired overlap between a dataset of natural language phrases, which is associated with a certain category, and a training dataset configured for training a machine learning model (So: § 1, pp 1, 2; § 7, pp 7; § 9, pp 8; Table 1: system seeks to generate or establish a desired, high, etc. coverage level upon a dataset; “adding gazetteer features does not hurt the performance of the neural systems, but only improves it when the gazetteer has high coverage,” such that the inclusion of a gazetteer improves the neural network system with respect to the dataset “with enough coverage on the dataset, gazetteer features improve neural NER systems,” in So the value is implicit, high enough to provide sufficient coverage) wherein the certain category corresponds to the So entity type of a gazetteer which is a GPE, PER, etc. gazetteer; “entity names … organized by entity type,”);
determining a second coverage value by determining a number of natural language phrases from the dataset of natural language phrases that are present in the training dataset and associated in the training dataset with a gold label category matching the certain category (So: ¶ 3.2, pp 3, § 4.1, 3, 4; § 7, pp 6, 7; Table 1, 6: such as by computing by particular gazetteer training, dataset coverage per category, such as by matching gazetteer phrases to training entities of the gold label, entities of a given type, or category as used to determine the coverage values per type, category, etc. shown in table 6); and selecting from the dataset of natural language phrases, the one or more natural language phrases that are not present in the training dataset (So: § 8, pp 7, such as to provide or enhance the system with phrases supplying entities, names, etc. that the training data lacks; “drawing on gazetteer data may help introduce new patterns not present in the training data, such as ORGs beginning with "Association of",” which forms additional data to enhance the training data with respect to the coverage lacked in the training data); such that a tuned or retrained machine learning model (So: § 1, pp 1; § 4.2, pp 4; model tuned or retrained with respect to gazetteer data) processes contextual features generated from the natural language query and expressional features generated from the dataset of natural language phrases (So: § 4, pp 4; § 5, pp 5; Fig 3, 5: contextual features such as BERT embeddings concatenated with expressional features such as the gazetteer one hot vectors to improve the underlying model).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the instant application to utilize the gazetteer vs training coverage measurement and gazetteer driven training augmentation of So with the chatbot NLU pipeline, per-slot gazetteer, based generative grammar of Go for at least the purpose of fine tuning based on labelled training utterances and gold labelled intents to thereby improve an intent/skill classifier output by using a coverage metric to determine under resourced gazetteer conditions and remedy by generation and use of the Go taught synthetic data to improve coverage and yield a more accurate classifier; one of ordinary skill in the art would have expected only predictable results therefrom.
Go in view of So does not explicitly teach the explicit conditional control claimed such that if a coverage value is insufficient that a desired coverage value the system operates to modify the training dataset, retrain and deploy the model.
IN a related field of endeavor Ka teaches a system and method for improving a machine learning model comprising receiving a first coverage value (Ka: Abstract; Col 13:52-13:67, 19:38-19:49; Claims 1, 2, 14, 18: system determines a value for a minimum coverage threshold); in response to determining that the second coverage value; is less than the first coverage value, modifying the training dataset to include one or more natural language phrases from the dataset of natural language phrases (Ka: Col 13:52-13:67, 19:1-19:8, 19:38-19:49; Claims 1, 2, 14, 18: when measured coverage fails to meet minimum threshold coverage additional training data is sources and the model iteratively trained); such as by retraining the machine learning model using the modified train dataset (Ka: Col 13:52-13:67, 19:1-19:8, 19:38-19:49; Claims 1, 2, 14, 18: system tunes or retrains based on the additional training data).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the instant application to utilize the Ka taught comparison of coverage values to initiate the generation of additional data and subsequent tuning or retraining of the Go in view of So system and method for at least the purpose of optimizing retraining based on determination of sufficiently high coverage in a gazetteer, improving the responses of a chatbot thereby, etc.; one of ordinary skill in the art would have expected only predictable results therefrom.
Regarding claim 3
Go in view of So in view of Ka teaches or suggests:
The computer-implemented method of claim 1, further comprising: prior to the processing, modifying the dataset of natural language phrases by adding, to the dataset of natural language phrases from the training dataset, one or more natural language phrases that are associated with the gold label category and not present in the dataset of natural language phrases, wherein, subsequent to modification of the dataset of natural language phrases, the modified dataset of natural language phrases includes a number of natural language phrases that are present in the training dataset in a proportion greater than or equal to the first coverage value (Go: § 3, pp 2; § 4.1, pp 3, 4; § 4.2, pp 4; § 5.1, pp 5: generating synthetic data comprising labelled training utterances); (So: § 4.1, pp 4; § 9, pp 7, 8: system contemplates such modifications to the gazetteer, categorical list, etc.); (Ka: Col 13:52-13:67, 19:1-19:8, 19:38-19:49; Claims 1, 2, 14, 18: such as by iteratively augmenting data, retraining based thereon and in response to values failing to meet a minimum threshold). The claim is considered obvious over Go as modified by So and Ka as addressed in the base claim as it would have been obvious to apply the further teaching of Go, So, and/or Ka to the modified device of Go, So, and Ja; one of ordinary skill in the art would have expected only predictable results therefrom.
Regarding claim 4
Go in view of So in view of Ka teaches or suggests:
The computer-implemented method of claim 1, wherein subsequent to modification of the dataset of natural language phrases includes a number of natural language phrases that are present in the modified training dataset in a proportion greater than or equal to the first coverage value (So: ¶ 3.2, pp 3, § 4.1, 3, 4; § 7, pp 6, 7; Table 1, 6: such as by computing by particular gazetteer training, dataset coverage per category, such as by matching gazetteer phrases to training entities of the gold label, entities of a given type, or category as used to determine the coverage values per type, category, etc. shown in table 6); (Ka: Col 13:52-13:67, 19:1-19:8, 19:38-19:49; Claims 1, 2, 14, 18: such as by iteratively augmenting data, retraining based thereon and in response to values failing to meet a minimum threshold). The claim is considered obvious over Go as modified by So and Ka as addressed in the base claim as it would have been obvious to apply the further teaching of Go, So, and/or Ka to the modified device of Go, So, and Ja; one of ordinary skill in the art would have expected only predictable results therefrom.
Regarding claim 8
Par in view of Ja teaches or suggests:
The computer-implemented method of claim 1, wherein: the contextual features and the expressional features are included in a set of input features (Go: § 4.1, pp 2, 3; Fig 1: such as using the model fine-tuned with the additional, replaced, etc. features by concatenating contextual features such as word/embedding type features with expressional features from the categorical list or sampled categorical list data); (So: § 4, pp 4; § 5, pp 5; Fig 3, 5: contextual features such as BERT embeddings concatenated with expressional features such as the gazetteer one hot vectors to improve the underlying model), the machine learning model is a convolutional neural network, and the set of input features corresponds to input nodes of the convolutional neural network (So: § 1, pp 1; § 5, pp 4, 5; Fig 5: system uses CNCN and Bi-LSTM to receive training data, user data; gazetteer data as input).
Regarding claim 9, 15—the claim is considered to recite substantially similar subject matter to that of claim 1 and is similarly rejected.
Regarding claim 11, 17—the claim is considered to recite substantially similar subject matter to that of claim 3 and is similarly rejected.
Regarding claim 12, 18—the claim is considered to recite substantially similar subject matter to that of claim 4 and is similarly rejected.
Regarding claim 21
Go in view of So in view of Ka teaches or suggests:
The computer-implemented method of claim 1 further comprising: causing the chatbot system to select the identified skill or intent for responding to the natural language query (Go: § 3, 4.1: system determines intents corresponding to user intention, finds skills such as in the ASK kit corresponding to custom functionality based thereon); (Ka: Col 5:65-6:5, 8:44-8:64: classification directs response selection based on a determined context). The claim is considered obvious over Go as modified by So and Ka as addressed in the base claim as it would have been obvious to apply the further teaching of Go, So, and/or Ka to the modified device of Go, So, and Ja; one of ordinary skill in the art would have expected only predictable results therefrom.
Response to Arguments
Applicant’s arguments in concert with claim amendments, see Remarks and Claims, filed 5/26/26, with respect to the rejection(s) of claim(s) 1-20 under 35 USC 103 over Parada San Martin in view of Jacob have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made of claims 1, 3, 4, 8, , 9, 11, 12, 15, 18, 21 under 35 USC 103 over Goyal in view of Song in view of Kang.
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 PAUL C MCCORD whose telephone number is (571)270-3701. The examiner can normally be reached 730-630 M-F.
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, CAROLYN EDWARDS can be reached at (571) 270-7136. 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.
/PAUL C MCCORD/Primary Examiner, Art Unit 2692
Prosecution Insights