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 1-20 are pending. Claims 1, 6, 11, and 16 are independent.
This Application was published as US 20240256789.
Apparent priority is 15 October 2021.
The instant Application is directed to a method of responding to queries by determining the dialog type and inputting the dialog type along with the dialog into a response generation network.
Claim Objections
Claim 5 objected to because of the following informalities: line 4, “at last one” is understood to mean “at least one”. Appropriate correction is required.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1-3, 5-8, 10-13, 15-18, and 20 is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Liang (US 20220083742 A1).
Regarding claim 1, Liang discloses: 1. A response determining method, wherein the method comprises: obtaining a to-be-responded first user statement; ("[0083] The obtaining module is configured to obtain a current dialogue sentence input by a user." )
determining first state information of the first user statement based on the first user statement by using a state determining network, ("[0086] The first neural network module is configured to: use the current dialogue sentence and a goal type and a goal entity of a preceding dialogue sentence obtained before the current dialogue sentence as an input; generate a goal type and a goal entity of the current dialogue sentence through feature extraction." )
wherein the first state information comprises a first dialog type of the first user statement, and the first dialog type is a chit-chat dialog, a task- oriented dialog, a question answering dialog, or a retrieval dialog; and ("[0084] The neural network system is configured to: use the current dialogue sentence and knowledge base data as an input, and generate a reply sentence through feature extraction, where the reply sentence is a chitchat sentence, an answer sentence or a recommendation sentence." )
inputting the first user statement and the first dialog type into a response generation network, to obtain a response corresponding to the first user statement. ("[0087] The second neural network module is configured to: use the current dialogue sentence, the goal type and the goal entity of the current dialogue sentence and the knowledge base data as an input; and generate the reply sentence through feature extraction and classification." )
Regarding claim 2, Liang discloses: 2. The method according to claim 1, wherein the determining first state information of the first user statement based on the first user statement by using a state determining network comprises: determining the first dialog type of the first user statement from a plurality of dialog types by using a state determining network, wherein the plurality of dialog types comprise at least two of the chit-chat dialog, the task-oriented dialog, the question answering dialog, or the retrieval dialog. ("[0084] The neural network system is configured to: use the current dialogue sentence and knowledge base data as an input, and generate a reply sentence through feature extraction, where the reply sentence is a chitchat sentence, an answer sentence or a recommendation sentence." )
Regarding claim 3, Liang discloses: 3. The method according to claim 1, wherein the method further comprises: obtaining a to-be-responded second user statement; determining second state information of the second user statement based on the second user statement by using the state determining network, wherein the second state information comprises a second dialog type of the second user statement, the second dialog type is a chit-chat dialog, a task-oriented dialog, a question answering dialog, or a retrieval dialog, and the second dialog type is different from the first dialog type; and inputting the second user statement and the second dialog type into the response generation network, to obtain a response corresponding to the second user statement. ("[0081] In summary, the man-machine dialogue method provided by the example may fuse various types of dialogues by the neural network, so as to actively and naturally guide the man-machine dialogues from non-recommendation dialogues such as chitchat dialogues, question answering dialogues, task-based dialogues to recommendation dialogues, and fuse the knowledge base data naturally into the dialogue; and the technical solution may accurately generate and output to the user the reply sentences including chitchat sentences, answering sentences and recommendation sentences by the neural network, and then realize the recommendation goal through one or more dialogue interactions in time and accurately on the basis of the knowledge base data and the user interest obtained by analyzing the dialogue sentences input by the user, which may enhance the initiative, scalability and richness of the man-machine dialogue, thereby enhancing the user experience." – Liang discloses that the system can output a plurality of replies to a plurality of types of queries.)
Regarding claim 5, Liang discloses: 5. The method according to claim 1, wherein the inputting the first user statement and the first dialog type into a response generation network, to obtain a response corresponding to the first user statement comprises: obtaining, from at last one of the first user statement or a database based on the first user statement, a keyword or a key sentence for constructing the response; and ("[0086] The first neural network module is configured to: use the current dialogue sentence and a goal type and a goal entity of a preceding dialogue sentence obtained before the current dialogue sentence as an input; generate a goal type and a goal entity of the current dialogue sentence through feature extraction." )
inputting the first user statement, the first dialog type, and the keyword or the key sentence into the response generation network, to obtain the response corresponding to the first user statement. ("[0087] The second neural network module is configured to: use the current dialogue sentence, the goal type and the goal entity of the current dialogue sentence and the knowledge base data as an input; and generate the reply sentence through feature extraction and classification." )
Regarding claim 6, Liang discloses: 6. A response determining method, wherein the method comprises: obtaining a first user statement, a first dialog type of the first user statement, and a first response corresponding to the first user statement, wherein the first dialog type is a real type of the first user statement, and the first dialog type is a chit-chat dialog, a task-oriented dialog, a question answering dialog, or a retrieval dialog; determining first state information of the first user statement based on the first user statement by using a state determining network, wherein the first state information comprises a second dialog type of the first user statement; inputting the first user statement and the first dialog type into a response generation network, to obtain a second response corresponding to the first user statement; updating the state determining network based on a difference between the first dialog type and the second dialog type; and updating the response generation network based on a difference between the first response and the second response. (See mapping of claim 1. Liang further discloses training of the networks: "[0026] The training stage is used to determine, according to the training data, parameters of each network or model in the neural network system by maximizing a likelihood function on the training data through a back-propagation algorithm or a stochastic gradient descent algorithm. The usage stage is used to generate the reply sentence and return it to the user by having the current dialogue sentence input by the user as the input of the neural network system and performing calculations by the neural network system based on a knowledge base which has already been constructed." )
Regarding claim 7, Liang discloses: 7. The method according to claim 6, wherein the determining first state information of the first user statement based on the first user statement by using a state determining network comprises: determining the second dialog type of the first user statement from a plurality of dialog types by using the state determining network, wherein the plurality of dialog types comprise at least two of the chit-chat dialog, the task-oriented dialog, the question answering dialog, or the retrieval dialog. ("[0084] The neural network system is configured to: use the current dialogue sentence and knowledge base data as an input, and generate a reply sentence through feature extraction, where the reply sentence is a chitchat sentence, an answer sentence or a recommendation sentence." )
Regarding claim 8, Liang discloses: 8. The method according to claim 6, wherein the method further comprises: obtaining a second user statement, a third dialog type of the second user statement, and a third response corresponding to the second user statement, wherein the third dialog type is a real type of the second user statement; determining second state information of the second user statement based on the second user statement by using the state determining network, wherein the second state information comprises a fourth dialog type of the second user statement, and the fourth dialog type is different from the third dialog type; inputting the second user statement and the third dialog type into the response generation network, to obtain a fourth response corresponding to the second user statement; updating the state determining network based on a difference between the fourth dialog type and the third dialog type; and updating the response generation network based on a difference between the fourth response and the third response. (See claim 3. Liang further discloses multiple training samples: "[0033] The knowledge base is generated on the basis of some “facts”. The knowledge base includes records or “tuples”. Specifically, the knowledge base may be obtained from the Internet. In this example, the knowledge base data may be specifically composed of multiple triples. Web pages may be crawled from encyclopedia knowledge sites such as Baidu Encyclopedia, Interactive Encyclopedia, and Douban, and the structured triple may be obtained by analyzing a table in the web page. After further processing including denoising, merging and so on, multiple triples are extracted to form the knowledge base" - see also "[0024] obtaining the neural network system by performing training on training data, where the training data includes dialogue sequences, candidate reply sentences, the knowledge base data, and a target recommendation sequence, and the dialogue sequences include a chitchat dialogue, a question answering dialogue, and a recommendation dialogue." )
Regarding claim 10, Liang discloses: 10. The method according to claim 6, wherein the inputting the first user statement and the first dialog type into a response generation network, to obtain a second response corresponding to the first user statement comprises: obtaining, from the first user statement or a database based on the first user statement, a keyword or a key sentence for constructing the response; and inputting the first user statement, the first dialog type, and the keyword or the key sentence into the response generation network, to obtain the second response corresponding to the first user statement. ("[0086] The first neural network module is configured to: use the current dialogue sentence and a goal type and a goal entity of a preceding dialogue sentence obtained before the current dialogue sentence as an input; generate a goal type and a goal entity of the current dialogue sentence through feature extraction. [0087] The second neural network module is configured to: use the current dialogue sentence, the goal type and the goal entity of the current dialogue sentence and the knowledge base data as an input; and generate the reply sentence through feature extraction and classification." )
Claim 11 is an apparatus claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale. Additionally, at least one processor; and one or more memories of the Claim are taught by Liang ("CPU"; "RAM"; "ROM" Fig. 5).
Claim 12 is an apparatus claim with limitations corresponding to the limitations of Claim 2 and is rejected under similar rationale.
Claim 13 is an apparatus claim with limitations corresponding to the limitations of Claim 3 and is rejected under similar rationale.
Claim 15 is an apparatus claim with limitations corresponding to the limitations of Claim 5 and is rejected under similar rationale.
Claim 16 is an apparatus claim with limitations corresponding to the limitations of Claim 6 and is rejected under similar rationale. Additionally, at least one processor; and one or more memories of the Claim are taught by Liang ("CPU"; "RAM"; "ROM" Fig. 5).
Claim 17 is an apparatus claim with limitations corresponding to the limitations of Claim 7 and is rejected under similar rationale.
Claim 18 is an apparatus claim with limitations corresponding to the limitations of Claim 8 and is rejected under similar rationale.
Claim 20 is an apparatus claim with limitations corresponding to the limitations of Claim 10 and is rejected under similar rationale.
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.
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.
Claim(s) 4, 9, 14, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liang in view of Dymetman et al. (US 20220108081 A1).
Regarding claim 4, Liang does not disclose the additional limitations.
Dymetman discloses: 4. The method according to claim 1, wherein the state determining network and the response generation network each are a generative pre-trained transformer (GPT) model, a dialogue generative pre-trained transformer (DialoGPT) model, a bidirectional and auto-regressive transformer (BART) model, or a transfer text-to-text transformer (T5) model. ("[0006] Language models (LMs), in a strict sense, such as GPT-2 and GPT-3, or in an extended sense, such as BERT, pre-trained on large datasets of text, are well known in natural language processing. Such pre-trained language models are typically seen as stores of generic knowledge that can be used for downstream tasks through fine-tuning, often done by providing a small amount of task-specific training data and extending or modifying parameters of pre-trained language models..." )
Liang and Dymetman are considered analogous art to the claimed invention because they disclose methods of text generation. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Liang with a GPT model for the neural networks as disclosed by Dymetman. Doing so would have been beneficial because it has stores of generic knowledge. (Dymetman [0006]) This combination falls under combining prior art elements according to known methods to yield predictable results or simple substitution of one known element for another to obtain predictable results. See MPEP 2141, KSR, 550 U.S. at 418, 82 USPQ2d at 1396.
Regarding claim 9, Liang does not disclose the additional limitations.
Dymetman discloses: 9. The method according to claim 6, wherein the state determining network and the response generation network each are a generative pre-trained transformer (GPT) model, a dialogue generative pre-trained transformer (DialoGPT) model, a bidirectional and auto-regressive transformer (BART) model, or a transfer text-to-text transformer (T5) model. ("[0006] Language models (LMs), in a strict sense, such as GPT-2 and GPT-3, or in an extended sense, such as BERT, pre-trained on large datasets of text, are well known in natural language processing. Such pre-trained language models are typically seen as stores of generic knowledge that can be used for downstream tasks through fine-tuning, often done by providing a small amount of task-specific training data and extending or modifying parameters of pre-trained language models..." )
See claim 4 for motivation statement.
Claim 14 is an apparatus claim with limitations corresponding to the limitations of Claim 4 and is rejected under similar rationale.
Claim 19 is an apparatus claim with limitations corresponding to the limitations of Claim 9 and is rejected under similar rationale.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Wang et al. (US 20220414737 A1). Wang discloses that GPT, DialoGPT, BART, and T5 models are known machine learning models for processing queries. (See [0038])
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JON C MEIS whose telephone number is (703)756-1566. The examiner can normally be reached Monday - Thursday, 8:30 am - 5:30 pm EST.
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, Hai Phan can be reached at 571-272-6338. 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.
/JON CHRISTOPHER MEIS/Examiner, Art Unit 2654
/HAI PHAN/Supervisory Patent Examiner, Art Unit 2654