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 .
Remarks
This action is reopened. The previous action filed 07/07/2026 is replaced/superseded. This non-final rejection is sent to include 35 USC 101 software per se rejections for claim 15, and include new rejection for claims 4, and 9, because Hu's date not did not precede the application's EFD, as well as objections to claim 12.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 11/06/2024 and 06/22/2023 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claim 12 objected to because of the following informalities:
Claim 12 line 1 recites the limitation, “A computer readable and non-transitory recording medium”. The examiners note that a “non-transitory recording medium” is statutory, but a “A computer readable recording medium” is not. This may be corrected by deleting the “and” within the recited limitation. Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 15 is also rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because the claimed product is software-per-se, as the program needs to be embodied a computer readable medium for its functionality to be realized. Claim 15 recites, “A machine learning model capable of discriminating whether article information having a plurality of modalities is genuine or not…”. Applicant’s claim language recites models including discriminators with no further structure to indicate that the model is anything other than software. Products do not have a physical or tangible form, such as a computer program per se (often referred to as "software per se") when claimed as a product without any structural recitations. As the courts' definitions of machines, manufactures and compositions of matter indicate, a product must have a physical or tangible form in order to fall within one of these statutory categories. Digitech, 758 F.3d at 1348, 111 USPQ2d at 1719. Thus, the claim fails to fall within a statutory category. The claims may be amended to narrow the claim to cover only statutory embodiments to avoid a rejection under 35 U.S.C. 101. Claim Rejections - 35 USC § 102 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.
Claim 16-17 are dependent upon Claim 1, incorporate the deficiencies of claim 1, and do not add tangibility to the claimed subject matter, they are likewise rejected.
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.
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.
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.
Claim(s) 1-3, 5, 7-8, & 11-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20190236614 “Burgin” in light of Ben-Yosef, M., & Weinshall, D. (2018). Gaussian Mixture Generative Adversarial Networks for Diverse Datasets, and the Unsupervised Clustering of Images. ArXiv.org. https://arxiv.org/abs/1808.10356, hereinafter “Ben-Yousef”.
Claim 1:
Burgin teaches a detecting device comprising:
a second machine learning model provided for mutual learning together with a first machine learning model, the first machine learning model being subjected to machine learning so as to generate genuine or fraudulent article information having a plurality of modalities, and the second machine learning model being subjected to machine learning so as to discriminate whether article information having a plurality of modalities is genuine or not (i.e. para. [0038], “The Generative Adversarial Network (GAN) subsystem 130, discussed in more detail with respect to FIG. 5, may include a generator that generates simulated counterfeit images and a discriminator that trains on these and/or images of actual fake items to classify input images as those of faked or genuine items”, wherein the BRI for article information having a plurality of modalities encompasses the how a photograph image with text may be an article of information, in which the generated images and associated information generated by a first generator MLM is ingested by a second discriminator MLM with an OCR module);
wherein the second machine learning model includes (i.e. para. [0028], “The system may use advanced convolutional neural network (CNN) models tuned to first detect a counterfeit good and unique preprocessing used to perform fake classification. Various techniques, such as material texture, pattern recognition, precise object geometry, and object feature position, may be utilized by the system. In an example, the system may identify fake goods, such as fake medicine products using supervised learning”, wherein the it is noted that a plurality of techniques are evaluated and combined for each modality of texture, pattern, and OCR and are used in combination when determining the probability of a fake classification), and the second machine learning model is configured to output a probability that the article information is genuine based on outputs (i.e. para. [0061], “ the discriminator 506 may generate an output based on convolutional processing of an input image of an item being undergoing counterfeit analysis, the output including a prediction of whether the item is a counterfeit. … The convolutional layers output 509 may each reflect an analysis of a respective portion of an input image”, wherein it is noted that the system is multimodal as it generates probabilities for each element (stitching, graphics, text, etc.) and passes them to a classifier which weights the various modalities to reach an overall determination of whether the item is genuine); at least one memory configured to store program code;
at least one processor configured to operate as instructed by the program code (i.e. para. [0071], The memory 110 may be, for example, Random Access memory (RAM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a storage device, an optical disc, and the like. The computer readable storage medium 1104 may be a non-transitory machine-readable storage medium, where the term “non-transitory” does not encompass transitory propagating signals), the program code including;
obtaining code configured to cause at least one of the at least one processor to obtain the article information having a plurality of modalities (i.e. para. [0047], The generator 502 may generate simulated counterfeit images 503, which may be used to train the discriminator 506. The discriminator 506, trained from the simulated counterfeit images 503 and/or actual counterfeit images 505 from the counterfeit images repository 504); and an estimating code configured to cause at least one of the at least one processor estimate whether the article information that is obtained and has the plurality of modalities is genuine or not, by using the second machine learning model (i.e. para. [0061], the discriminator 506 may generate an output based on convolutional processing of an input image of an item being undergoing counterfeit analysis, the output including a prediction of whether the item is a counterfeit).
While Burgin teaches using a discriminator using and evaluating plurality of modalities, Burgin may not explicitly teach
a plurality of discriminators, each discriminator corresponding to a respective modality of the article information,
the second machine learning model is configured to output a probability that the article information is genuine based on outputs of the plurality of discriminators.
However, Ben-Yosef teaches
a plurality of discriminators, each discriminator corresponding to a respective modality of the article information (i.e. pg. 4, “In the supervised setting, we change the GM-GAN’s discriminator so that instead of returning a single scalar, it returns a vector o ∈ RN where N is the number of classes in the dataset…. this modification can be thought of as having N binary discriminators, where each discriminator i is trained to separate real samples of class i from fake samples of class i and from real samples of classes other than class I”, wherein the BRI for a plurality of discriminators for each respective modality encompasses N binary discriminator for each class i),
the second machine learning model is configured to output a probability that the article information is genuine based on outputs of the plurality of discriminators (i.e. pg. 5, Here x denotes an image sampled from the Generator, p(y|x) denotes the inferred class label probability given x by the Inception network).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add a plurality of discriminators, each discriminator corresponding to a respective modality of the article information, the second machine learning model is configured to output a probability that the article information is genuine based on outputs of the plurality of discriminators, to Burgin’s system considers all the modalities when discriminating, with how a discriminator may be a plurality of discriminators for each modality, as taught by Ben-Yosef. One would have been motivated to combine Ben-Yosef with Burgin, and would have had a reasonable expectation of success in doing so as using techniques may improve the performance of GANs when the training dataset has large inter-class and intra-class diversity.
Claim 2:
Burgin and Ben-Yousef teach the detecting device according to claim 1,
wherein the first machine learning model is subjected to machine learning so as to generate fraudulent article information obtained by editing and processing at least part of genuine article information having a plurality of modalities (i.e. para. [0020], , the generator may take as input parameters from human users that direct the generator to make certain changes to a genuine image to create a simulated counterfeit image. For instance, the parameter may specify that a font recognized on a genuine image of a pill be adjusted to create a simulated counterfeit image of the pill).
Claim 3:
Burgin and Ben-Yousef teach the detecting device according to claim 1,
wherein the article information having a plurality of modalities includes image data representing an article image (i.e. para. [0035], supervised learning is done through a database maintained containing item information such as description and unique identifier information. This is linked to photographs and text examples and explanations of counterfeiting of that item. This is in addition to images and text descriptions of the genuine item).
Claim 5:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
Burgin further teaches
wherein the article information having a plurality of modalities includes text data representing an article descriptive sentence (i.e. para. [0035], “a database maintained containing item information such as description and unique identifier information. This is linked to photographs and text examples and explanations of counterfeiting of that item. This is in addition to images and text descriptions of the genuine item”, wherein the BRI for an article descriptive sentence encompasses a text description).
Claim 7:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
Burgin further teaches
wherein the article information having a plurality of modalities includes attribute data representing an article category (i.e. para. [0028], “the discriminator 506 may compare the features of the counterfeit images with the features identified from the known good images of the item. For example, the discriminator 506 may compare specific regions of an imaged item. Such regions may correspond to circularity or shape features of a pill, a partition of the pill, a specific attribute of the pill, and/or other features”, wherein the BRI for article category encompasses attribute information regarding the article of information such as the type or shape of medicine being analyzed).
Claim 8:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
Burgin further teaches
wherein the article information having a plurality of modalities includes numerical data representing a shipment timing (i.e. para. [0030], statistical models based on geography, shipping route, and frequency of fraud for an item can also generate a probability of fraud and feed into classification and probability determination techniques. Also, shipping route information such as start of item journey and intermediaries may also be used).
Claim 11:
Claim 11 is the method claim reciting similar limitations to claim 1 and is rejected for similar reasons.
Claim 12:
Claim 12 is the medium claim reciting similar limitations to claim 1 and is rejected for similar reasons.
Claim 13:
Claim 13 is the device claim reciting similar limitations to claim 1 and is rejected for similar reasons.
Claim 14:
Claim 14 is the method claim reciting similar limitations to claim 1 and is rejected for similar reasons.
Claim 15:
Claim 15 is the model claim reciting similar limitations to claim 1 and is rejected for similar reasons.
Claim 16:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
Burgin further teaches
wherein the first machine learning model includes a plurality of information generation sections, and each of the information generation sections generates information of a corresponding modality of the genuine or fraudulent article information (i.e. para. [0038], “Each convolutional layer output may reflect a weighted output that is based on a relative importance of a decision from each convolutional layer. The GAN subsystem 130 may further integrate an explainable AI feature in which the reasons for the classification may be interrogated by a classification activation module and used to solicit feedback from subject matter experts. The feedback may indicate whether or not various weighted decisions were appropriate”, wherein the various section process information across different modalities for both genuine and fraudulent items in order to produced data necessary for the system to determine authenticity).
Claim 17:
Burgin and Ben-Yousef teach the detecting device according to claim 16.
Burgin further teaches
Ben-Yousef further teaches
wherein the discriminators of the second machine learning model are configured to output a probability that information of the respective modality is genuine, for the information generated by the respective corresponding information generation sections (i.e. pg. 5, Here x denotes an image sampled from the Generator, p(y|x) denotes the inferred class label probability given x by the Inception network).
Claim(s) 4 & 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over and further in light of U.S. Patent Application Publication NO. 20190236614 “Burgin”, in light of Ben-Yosef, M., & Weinshall, D. (2018). Gaussian Mixture Generative Adversarial Networks for Diverse Datasets, and the Unsupervised Clustering of Images. ArXiv.org. https://arxiv.org/abs/1808.10356, hereinafter “Ben-Yousef”, and further in light of Zellers, R., Holtzman, A., Rashkin, H., Bisk, Y., Farhadi, A., Roesner, F., Choi, Y., & Allen, P. (2020, December 11). Defending Against Neural Fake News. https://arxiv.org/pdf/1905.12616, hereinafter “Zellers”.
Claim 4:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
However, Burgin may not explicitly teach
wherein the article information having a plurality of modalities includes text data representing an article title.
However, Zellers teaches
wherein the article information having a plurality of modalities includes text data representing an article title (i.e. [pg. 3, Section 3: Grover: Modeling Conditional Generation of Neural Fake News, para. 3], “However, a news article has necessary structure beyond the running text, or body field. Metadata fields include the domain where the article is published (indirectly marking the style), the date of publication, the names of the head authors, and line of the article itself.”, wherein the BRI for a plurality of modalities encompasses how a GAN considers text metadata such as an article headline).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the article information having a plurality of modalities includes text data representing an article title, to Burgin-Ben-Yousef’s system that generates articles with a plurality of information including text data, a generator may intake real and fake article titles and generate representative article data in order to take a GAN or CNN-based approach to training a model to detect and differentiate fake and genuine news articles, as taught by Zellers. One would have been motivated to combine Zellers with Burgin-Ben-Yousef, and would have had a reasonable expectation of success in doing so as the combination when used in a discriminative setting, represents an exciting opportunity for defense against neural fake news: the best models for generating neural disinformation are also the best models at detecting it.
Claim 9:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
Burgin may not explicitly teach
wherein the article information having a plurality of modalities includes attribute data representing an attribute of an exhibitor or a seller.
However, Zellers teaches
wherein the article information having a plurality of modalities includes attribute data representing an attribute of an exhibitor or a seller (i.e. [pg. 3, Section 3: Grover: Modeling Conditional Generation of Neural Fake News, para. 3], “However, a news article has necessary structure beyond the running text, or body field. Metadata fields include the domain where the article is published (indirectly marking the style), the date of publication, the names of the head authors, and line of the article itself.”, wherein the BRI for a plurality of modalities encompasses how a GAN considers text metadata such as a publisher or author)).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the article information having a plurality of modalities includes attribute data representing an attribute of an exhibitor or a seller, to Burgin-Ben-Yousef’s system that generates articles with a plurality of information including text data, with how a discriminator may user attribute data of an exhibitor publisher to detect and differentiate fake and genuine news articles, as taught by Zellers. One would have been motivated to combine Zellers with Burgin-Ben-Yousef, and would have had a reasonable expectation of success in doing so as the combination when used in a discriminative setting, represents an exciting opportunity for defense against neural fake news: the best models for generating neural disinformation are also the best models at detecting it.
Claim(s) 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over in light of U.S. Patent Application Publication NO. 20190236614 “Burgin”, in light of Ben-Yosef, M., & Weinshall, D. (2018). Gaussian Mixture Generative Adversarial Networks for Diverse Datasets, and the Unsupervised Clustering of Images. ArXiv.org. https://arxiv.org/abs/1808.10356, hereinafter “Ben-Yousef”, and further in light of U.S. Patent Application Publication NO. 20210125031 “Keng”.
Claim 6:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
Burgin may not explicitly teach
wherein the article information having a plurality of modalities includes numerical data representing an article price.
However, Keng teaches
wherein the article information having a plurality of modalities includes numerical data representing an article price (i.e. para. [0051], “Other approaches generate plausible customer e-commerce orders for a given product using a Generative Adversarial Network (GAN). Given a product embedding, some approaches generate a tuple containing a product embedding, customer embedding, price, and date of purchases, which summarizes a typical order. This approach using a GAN can provide insights into product demand, customer preferences, price estimation and seasonal variations by simulating what are likely potential order”, wherein the G aims to produce realistic samples from this distribution while a discriminator D tries to differentiate fake samples from real samples. By alternating optimization steps between the two components, the generator ultimately learns the distribution of the real data).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the article information having a plurality of modalities includes numerical data representing an article price, to Burgin- Ben-Yousef’s system that generates articles with a plurality of information in order to discriminate real and fake items, with how a GAN may use numerical price data differentiate real and fake articles of data, as taught by Keng. One would have been motivated to combine Keng with Burgin-Ben-Yousef, and would have had a reasonable expectation of success in doing so as the additional consideration of price results in better generated samples that are more similar to a real data distribution.
Claim(s) 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over in light of U.S. Patent Application Publication NO. 20190236614 “Burgin”, in light of Ben-Yosef, M., & Weinshall, D. (2018). Gaussian Mixture Generative Adversarial Networks for Diverse Datasets, and the Unsupervised Clustering of Images. ArXiv.org. https://arxiv.org/abs/1808.10356, hereinafter “Ben-Yousef”, and further in light of U.S. Patent Application Publication NO. 20210019339 “Ghulati”.
Claim 10:
Burgin and Ben-Yousef teach the detecting device according to claim 1.
Burgin may not explicitly teach
wherein the article information having a plurality of modalities includes numerical data representing an evaluation given to an exhibitor or a seller.
However, Ghulati teaches
wherein the article information having a plurality of modalities includes numerical data representing an evaluation given to an exhibitor or a seller (i.e. para. [0253], “a method and system for assessing the quality of content generated by a user and their position within a credibility graph in order to generate a reliable credibility score is provided. The credibility score may be determined for a person, organization, brand or piece of content by means of calculation using a combination of extrinsic signals, content signals… The method may be further capable of determining a credibility score of the user of the generated content by the combination of the score indicative of the credibility of the content and the score indicative of the credibility of the user”, wherein the BRI numerical data encompasses the credibility score associated with an article published by an organization).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the article information having a plurality of modalities includes numerical data representing an evaluation given to an exhibitor or a seller, to Burgin- Ben-Yousef’s system that generates articles with a plurality of information including text data, with how a discriminator may include intaking associate article information including a numerical data value representing a credibility given to an organization or person related to the article, as taught by Ghulati. One would have been motivated to combine Burgin-Ben-Yousef with Ghulati, and would have had a reasonable expectation of success in doing so as the additional consideration of credibility scoring may encourage the prevention of abuse and toxicity within online platforms.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
U.S. Patent Application Publication No. 20200184212 “Anthony” teaches in para. [0028], , the synthetic training documents generated in the previous step are used to train a fraud classification model. The fraud classification model may be any Al or machine learning model that takes one or more documents as input and outputs a classification for the one or more documents. The classification may identify the document(s) as potentially fraudulent or not.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DAVID H TAN whose telephone number is (571)272-7433. The examiner can normally be reached M-F 7:30-4:30.
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, Cesar Paula can be reached at (571) 272-4128. 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.
/D.T./Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights