DETAILED ACTION
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 .
Status of Claims
Claims 1-20 are pending and examined herein.
The specification is objected to.
Claims 1-20 are rejected under 35 U.S.C. 112(b).
Claims 1-20 are rejected under 35 U.S.C. 103.
Specification
The disclosure is objected to because of the following informalities: “general adversarial network” should likely be “generative adversarial network” in paragraphs [0004]-[0006].
Appropriate correction is required.
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.
Claims 1-20 are 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 applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 1, 8, and 15 recite the limitation “a general adversarial network (GAN)”. A GAN is a well-known acronym in machine learning meaning “generative adversarial network”. The specification states that a GAN is both a general adversarial network ([0004]-[0006]) and a generative adversarial network ([0024]-[0025]). Therefore, as it is not clear whether a GAN should mean “general adversarial network” or “generative adversarial network” the claim is rendered indefinite.
Dependent claims 2-7, 9-14, and 16-20 fail to resolve the issue and are rejected with the same rationale.
Claims 7 and 14 recite the limitation "the a difference between a pass rate of the first discriminative network and a pass rate of the second discriminative network". There is insufficient antecedent basis for this limitation in the claim.
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.
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.
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) 1-3, 5-6, 8-10, 12-13, 15-17, and 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma (US 2022/0180190 A1) and Peskov (“Multi-Domain Goal-Oriented Dialogues (MultiDoGO): Strategies toward Curating and Annotating Large Scale Dialogue Data”, 2019).
Regarding claim 1, Verma teaches
A computer-implemented method for generative artificial intelligence (Al), the method comprising: ([0009] states "Embodiments described herein provide methods, systems, and apparatuses for adapted generative adversarial networks for classification, including analyzing and training an adapted GAN with an additional module layer/network module referred to as the Head Discriminator (HD). For each class, a Generator (G), a Discriminator (D) and a Head Discriminator are trained." [0058] states "The various methods or processes outlined herein can be coded as software that is executable on one or more processors that employ any one of a variety of operating systems or platforms. Additionally, such software can be written using any of a number of suitable programming languages and / or programming or scripting tools, and also can be compiled as executable machine language code or intermediate code that is executed on a framework or virtual machine.")
receiving, via one or more processors, an input statement; and ([0026] states "For example an adapted GAN can be trained such that a bot can provide a weather forecast. Given the user message “Do I need my umbrella today?" the adapted GAN might classify the intent as “question about today's weather.” The chat bot might answer with the weather forecast for today by choosing a response, such as “Yes, you need an umbrella,” “It is going to rain,” or “It is rainy today” provided/selected by the response generation component." The user message is interpreted as the input statement. As the processor implements the method, the input statement must have been received by the processor.)
generating, via the one or more processors, a response to the input statement by inputting the input statement into a general adversarial network (GAN), the GAN comprising: ([0026] states "For example an adapted GAN can be trained such that a bot can provide a weather forecast. Given the user message “Do I need my umbrella today?" the adapted GAN might classify the intent as “question about today's weather.” The chat bot might answer with the weather forecast for today by choosing a response, such as “Yes, you need an umbrella,” “ It is going to rain, ” or “It is rainy today” provided/selected by the response generation component." As the GAN classifies the intent, the input statement must have been input into the GAN. As the chat bot responds, a response is generated. As the processor implements the method, the processor generates the response.)
a generative network configured to send and receive data to a discriminative network; and ([0020] states "Embodiments described herein provide methods, systems, and apparatuses for adapted Generative Adversarial Network (GAN) for classification, including analyzing and training an adapted GAN with an additional module layer / network module referred to as the Head Discriminator (HD). According to some embodiments, for each class, a Generator (G), a Discriminator (D) and a Head Discriminator are trained." [0029] states "For every iteration, the Discriminator can be trained on true data and output of the Generator, while the Generator is fed noisy and/or augmented data." Therefore, the discriminator receives data from the generative network. [0032] states "As discussed above, the Discriminator and the Generator can be trained as adversaries such that improvements in Discriminator output feeds back to improve the ability of the Generator to simulate the true data." Therefore, the discriminator sends data to the generative network.)
the discriminative network configured to send and receive data to the generative network, wherein the discriminative network was trained based on information of at least one domain, ([0029] states "For every iteration, the Discriminator can be trained on true data and output of the Generator, while the Generator is fed noisy and/or augmented data." Therefore, the generator sends data to the discriminator. [0032] states "As discussed above, the Discriminator and the Generator can be trained as adversaries such that improvements in Discriminator output feeds back to improve the ability of the Generator to simulate the true data." Therefore, the generator receives data from the generative network. [0030] states "For every iteration, the Discriminator can be trained on true data and output of the Generator, while the Generator is fed noisy and / or augmented data. The Generator can be trained on noisy and/or augmented data." As all data has a domain, the discriminator was trained based on information of at least one domain, and the class of the of the first discriminator is interpreted as the first domain.)
Verma does not appear to explicitly teach
wherein the at least one domain includes at least one of: retirement; cyber; legal; compliance; human resources; privacy; or fairness.
However, Peskov—directed to analogous art—teaches
wherein the at least one domain includes at least one of: retirement; cyber; legal; compliance; human resources; privacy; or fairness. (Page 4529 states, in reference to the included domains, that "Software domain involves customers inquiring about software services: products, outages, promotions, and bills." This is interpreted as the cyber domain.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Verma and Peskov because, as Peskov states on page 4526, "Moreover, recent work (Dong et al., 2015; Devlin et al., 2018) has shown the benefit of applying joint-training and transfer learning techniques to natural language processing tasks. How ever, these approaches have yet to become widely used in dialogue tasks, due to a lack of large scale datasets. Furthermore, the latest state of the art end-to-end neural approaches benefit from such training data even more so than past work on goal-oriented dialogues structured around slot filling (Lemon et al., 2006; Wang and Lemon, 2013)."
Regarding claim 2, the rejection of claim 1 is incorporated herein. Verma teaches
wherein: (i) the input statement includes a question or a request for information, and ([0026] states "Given the user message ‘Do I need my umbrella today?’ the adapted GAN might classify the intent as ‘question about today's weather.’” Therefore, the input statement includes a question/request for information.)
(ii) the generated response includes an answer to the question or a response to the request for more information. ([0026] further states "The chat bot might answer with the weather forecast for today by choosing a response, such as ‘Yes, you need an umbrella,’ ‘It is going to rain,’ or ‘It is rainy today’ provided/selected by the response generation component." Therefore, the response includes an answer to the question/response to the request for more information.)
Regarding claim 3, the rejection of claim 1 is incorporated herein. Verma teaches
training, via the one or more processors, the discriminative network by inputting the information of the at least one domain into the discriminative network. ([0030] states "For every iteration, the Discriminator can be trained on true data and output of the Generator, while the Generator is fed noisy and / or augmented data. The Generator can be trained on noisy and/or augmented data." As the discriminator is trained by the true data, the information of the at least one domain is input into the discriminator.)
Regarding claim 5, the rejection of claim 1 is incorporated herein. Verma teaches
displaying, via the one or more processors, on a display, the generated response. ([0026] states "The adapted GAN can be used to classify the intent of some user input (received, e.g., from a user communication device) and give an answer using a response generation component for classified class. For example an adapted GAN can be trained such that a bot can provide a weather forecast. Given the user message “Do I need my umbrella today?" the adapted GAN might classify the intent as “question about today's weather.” The chat bot might answer with the weather forecast for today by choosing a response, such as “Yes, you need an umbrella,” “It is going to rain,” or “It is rainy today” provided/selected by the response generation component." One of ordinary skill in the art would realize, as the user communication device is used to receive a user input, that the response would be displayed on the user communication device in order for the user to receive the response.)
Regarding claim 6, the rejection of claim 1 is incorporated herein. Verma teaches
the discriminative network is a first discriminative network; ([0036] states "FIG. 9 illustrates a novel class detector, according to an embodiment. GANs described herein are operable to assess a data input (labeled “ unknown input ”) and determine whether the class of the data input is associated with a previously recognized class (For example, associated with a previously trained Generator/Discriminator pair). The data input can be fed into each previously trained Generator/Head Discriminator pairs. Each Generator/Head Discriminator pair can be associated with a different data class." The discriminator associated with the first class is interpreted as a first discriminative network.)
the at least one domain is a first at least one domain; and (The class of the first discriminative network is interpreted as the first at least one domain.)
the GAN further comprises a second discriminative network, wherein the second discriminative network was trained based on information of a second at least one domain, wherein the second at least one domain: (Fig. 9 shows more discriminative networks, each numbered with a class. The class is interpreted as the domain, meaning that there are two at least one domains.)
is different than the first at least one domain, and (The different classes are interpreted as the different domains.)
Verma does not appear to explicitly teach
includes at least one of: retirement; cyber; legal; compliance; human resources; privacy; or fairness.
However, Peskov—directed to analogous art—teaches
includes at least one of: retirement; cyber; legal; compliance; human resources; privacy; or fairness. (Page 4529 states, in reference to the included domains, that "Finance domain simulates dialogues a customer may have with a bank. These include opening a bank account, checking their balance, and reporting a lost credit card." This is interpreted as a privacy domain, as the dialogues include intents relating to privacy, i.e. opening a bank account, checking their balance, and reporting a lost credit card.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Verma and Peskov for the reasons given above in regards to claim 1.
Regarding claim 8, Verma teaches
A system for generative artificial intelligence (Al), comprising one or more processors configured to: ([0009] states "Embodiments described herein provide methods, systems, and apparatuses for adapted generative adversarial networks for classification, including analyzing and training an adapted GAN with an additional module layer/network module referred to as the Head Discriminator (HD). For each class, a Generator (G), a Discriminator (D) and a Head Discriminator are trained." [0058] states "The various methods or processes outlined herein can be coded as software that is executable on one or more processors that employ any one of a variety of operating systems or platforms. Additionally, such software can be written using any of a number of suitable programming languages and / or programming or scripting tools, and also can be compiled as executable machine language code or intermediate code that is executed on a framework or virtual machine."
The remainder of claim 8 recites substantially similar subject matter to claim 1 and is rejected with the same rationale, mutatis mutandis.
Claims 9-10 and 12-13 recite substantially similar subject matter to claims 2-3 and 5-6 respectively and are rejected with the same rationale, mutatis mutandis.
Regarding claim 15, Verma teaches
A computer device for generative artificial intelligence (Al), the computer device comprising: ([0009] states "Embodiments described herein provide methods, systems, and apparatuses for adapted generative adversarial networks for classification, including analyzing and training an adapted GAN with an additional module layer/network module referred to as the Head Discriminator (HD). For each class, a Generator (G), a Discriminator (D) and a Head Discriminator are trained." [0058] states "The various methods or processes outlined herein can be coded as software that is executable on one or more processors that employ any one of a variety of operating systems or platforms. Additionally, such software can be written using any of a number of suitable programming languages and / or programming or scripting tools, and also can be compiled as executable machine language code or intermediate code that is executed on a framework or virtual machine."
one or more processors; and ([0058] states "The various methods or processes outlined herein can be coded as software that is executable on one or more processors that employ any one of a variety of operating systems or platforms.)
one or more non-transitory memories, the one or more non-transitory memories having stored thereon computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to: (To perform the method, a non-transitory memory must be present with computer-executable instructions that cause the processor to perform the method.)
The remainder of claim 15 recites substantially similar subject matter to claim 1 and is rejected with the same rationale, mutatis mutandis.
Claims 16-17 and 19-20 recite substantially similar subject matter to claims 2-3 and 5-6 respectively and are rejected with the same rationale, mutatis mutandis.
Claim(s) 4, 11, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma (US 2022/0180190 A1) and Peskov (“Multi-Domain Goal-Oriented Dialogues (MultiDoGO): Strategies toward Curating and Annotating Large Scale Dialogue Data”, 2019), and further in view of Steen (“A Gentle Introduction to Self-Training and Semi-Supervised Learning”, 2020).
Regarding claim 4, the rejection of claim 1 is incorporated herein. Verma does not appear to explicitly teach
training, via the one or more processors, the discriminative network
(i) in a first phase comprising a supervised training process, and
(ii) a second phase comprising an unsupervised training process.
However, Steen—directed to analogous art—teaches
training, via the one or more processors, the discriminative network (Page 2 states "This is where semi-supervised learning comes in. In taking a semi-supervised approach, we can train a classifier on the small amount of labeled data, and then use the classifier to make predictions on the unlabeled data.")
(i) in a first phase comprising a supervised training process, and (Page 3 states "Step 1: Split the labeled data instances into train and test sets. Then, train a classification algorithm on the labeled training data.")
(ii) a second phase comprising an unsupervised training process. (Page 3 states "Step 2: Use the trained classifier to predict class labels for all of the unlabeled data instances. Of these predicted class labels, the ones with the highest probability of being correct are adopted as ‘pseudo-labels’. (A couple of variations on Step 2: a) All of the predicted labels can be adopted as ‘pseudo-labels’ at once, without considering probability, or b) The ‘pseudo-labeled’ data can be weighted by confidence in the prediction.) Step 3: Concatenate the ‘pseudo-labeled’ data with the labeled training data. Re-train the classifier on the combined ‘pseudo-labeled’ and labeled training data." Thus, as the second phase uses the unlabeled data, the second phase is an unsupervised training process.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Verma and Peskov with the teachings of Steen, because as Steen states on page 2, "So, what if we only have enough time and money to label some of a large data set, and choose to leave the rest unlabeled? Can this unlabeled data somehow be used in a classification algorithm? This is where semi-supervised learning comes in. In taking a semi-supervised approach, we can train a classifier on the small amount of labeled data, and then use the classifier to make predictions on the unlabeled data. Since these predictions are likely better than random guessing, the unlabeled data predictions can be adopted as ‘pseudo-labels’ in subsequent iterations of the classifier."
Claims 11 and 18 recite substantially similar subject matter to claim 4 and are rejected with the same rationale, mutatis mutandis.
Claim(s) 7 and 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma (US 2022/0180190 A1) and Peskov (“Multi-Domain Goal-Oriented Dialogues (MultiDoGO): Strategies toward Curating and Annotating Large Scale Dialogue Data”, 2019), and further in view of Dines (US 2025/0054327 A1).
Regarding claim 7, the rejection of claim 6 is incorporated herein. Verma does not appear to explicitly teach
detecting, via the one or more processors, that the a difference between a pass rate of the first discriminative network and a pass rate of the second discriminative network is above a predetermined mismatch threshold; and
in response to the detecting that the difference between the pass rate of the first discriminative network and the pass rate of the second discriminative network is above the predetermined mismatch threshold, generate, via the one or more processors, an alert.
However, Dines—directed to analogous art—teaches
detecting, via the one or more processors, that the a difference between a pass rate of the first discriminative network and a pass rate of the second discriminative network is above a predetermined mismatch threshold; and ([0130] states "The CV and/or OCR model and the generative AI model process the respective source information that was provided to each model at 820 to extract data elements therein, and outputs from these models are analyzed by an RPA robot or another software application at 830 to determine whether the models agree on the extracted data element predictions. The model outputs may include extracted data element predictions, respective confidence scores, the extracted data elements themselves, etc. If the CV and/or OCR models agree on the prediction, and the CV and/or OCR models and the generative AI model identified all data elements in the source with at least their respective acceptance thresholds (whether common or specific to that model) at 840, the common extracted element predictions and/or the respective data elements are provided as output at 850." The models are interpreted as the discriminative networks and their predictions are interpreted as the pass rate. The threshold at which they are determined to disagree is interpreted as the predetermined mismatch threshold.)
in response to the detecting that the difference between the pass rate of the first discriminative network and the pass rate of the second discriminative network is above the predetermined mismatch threshold, generate, via the one or more processors, an alert. ([0131] states "However, if the models disagree on the prediction or the CV and/or OCR models and/or the generative AI model are below their respective acceptance thresholds regardless of model agreement at 840, human assistance is requested at 860 (i.e., human in the loop).")
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Verma and Peskov with the teachings of Dines because, as Dines states in [0131], "If the prediction was wrong, the correction in the labeled data may help the element to be successfully identified in the future after retraining."
Claim 14 recites substantially similar subject matter to claim 7 and is rejected with the same rationale, mutatis mutandis.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JESSICA THUY PHAM whose telephone number is (571)272-2605. The examiner can normally be reached Monday - Friday, 9 A.M. - 5:00 P.M..
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, Li Zhen can be reached at (571) 272-3768. 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.
/J.T.P./ Examiner, Art Unit 2121
/Li B. Zhen/ Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights