DETAILED ACTION
This action is in response to a filing filed on May 2nd, 2025. Claims 1-20 have been examined in this application. The Information Disclosure Statement (IDS) filed on May 2nd, 2025, and June 1st, 2026 has been acknowledged.
Priority
This application is a continuation of US Patent Application No. 17/084,701, filed October 30, 2020, which claims priority to US Patent Application No. 62/933,175, filed November 8, 2019. The disclosure of these applications are hereby incorporated by reference in their entireties.
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 .
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP §§ 706.02(l)(1) - 706.02(l)(3) for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/process/file/efs/guidance/eTD-info-I.jsp.
Claims 1-20 rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-5 of U.S. Patent No. 12,293,375. Although the claims at issue are not identical, they are not patentably distinct from each other because instant claim 1 is anticipated by the conflicting patented claim 1 as shown in the table below. The difference between the instant examined claim and the conflicting patented claim is that the conflicting patented claim is narrower in scope and falls within the scope of the examined claim. Thus, the species or sub-genus claimed in the conflicting patent anticipates the examined claimed genus. Therefore, a patent to the examined claim genus would improperly extend the right to exclude granted by a patent to the species or sub-genus should the genus issue as a patent after the species or sub-genus. See MPEP §804(II)(B)(1).
For reference, the following table matches the narrower limitations of method claim 1 of the patented parent application no. 17,084,701 (Pat. 12,293,375) with the similar limitations of method claim 1 of current child Application No. 19197452:
Claim 1 of Patented Application 12,293,375
Claim 1 of Application 19197452
A computer-implemented method of using machine learning to determine product certification eligibility, the method comprising:
A computer-implemented method of using machine learning to determine product certification eligibility, the method comprising:
training, by a computer processor, a plurality of machine learning models using a set of training data associated with a set of appliances, the set of training data comprising (i) textual and visual content corresponding to the set of appliances, wherein the visual content comprises, for each appliance of the set of appliances, a training schematic drawing depicting appliance wiring of that appliance, and wherein the training schematic drawing comprises a label for a category of a plurality of categories;
training, by a computer processor, a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, wherein the visual content comprises, for each product of the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories;
storing the plurality of machine learning models in a memory;
accessing, by the computer processor, a specification (i) comprising a schematic depicting an appliance wiring of an appliance, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the appliance, wherein the certification is applicable to the geographic area or jurisdiction;
storing the plurality of machine learning models in a memory;
accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction;
performing, by the computer processor, a visual analysis technique on the schematic to determine a set of components, a set of materials, and a length for the appliance wiring;
identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the appliance, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the appliance;
performing, by the computer processor, a visual analysis technique on the visual to determine a set of components and a set of materials associated with the product;
identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product;
analyzing, by the computer processor using the machine learning model, the specification, including the set of components, the set of materials, and the length for the appliance wiring to:
determine a set of keywords associated with the appliance, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication that the appliance is ineligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords;
And analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to:
determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords.
determining a set of changes that need to be implemented in the appliance that would result in the appliance being eligible for the certification;
automatically performing, by the computer processor without user intervention, a review to reconcile or clarify any of the set of keywords being the trigger keyword type;
and determining, by the computer processor, a reason when none of the set of keywords is the eligible keyword type.
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.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., an abstract idea) without significantly more.
Step 1: Claims 1-8 is/are drawn to method (i.e., a process), claims 9-14 is/are drawn to system (i.e., a manufacture), and claims 15-20 is/are drawn to computer-readable storage medium (i.e., a manufacture). As such, claims 1-20 is/are drawn to one of the statutory categories of invention (Step 1: YES).
Step 2A - Prong One: In prong one of step 2A, the claim(s) is/are analyzed to evaluate whether it/they recite(s) a judicial exception.
Representative Claim 1: A computer-implemented method of using machine learning to determine product certification eligibility, the method comprising:
training, by a computer processor, a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, wherein the visual content comprises, for each product of the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories;
storing the plurality of machine learning models in a memory;
accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction;
performing, by the computer processor, a visual analysis technique on the visual to determine a set of components and a set of materials associated with the product;
identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product;
and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to:
determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords.
(Examiner notes: The underlined claim terms above are interpreted as additional elements beyond the abstract idea and are further analyzed under Step 2A - Prong Two)
Under their broadest reasonable interpretation, the steps of training a plurality of machine learning models using training data associated with products, storing the machine learning models, accessing a product specification that includes a product visual, certification information, and a jurisdiction or geographic area, performing visual analysis on the visual to determine components and materials, selecting a machine learning model corresponding to the product and jurisdiction, analyzing the specification to determine keywords, classifying each keyword as ineligible, trigger, or eligible, and outputting an indication of whether the product is eligible for certification based on the keyword types. Under the broadest reasonable interpretation, the claim recites the abstract idea of evaluating product information under certification or regulatory rules to determine whether the product is eligible, ineligible, or requires further review. The claim recites collecting and analyzing product information, identifying product characteristics, applying classification rules or criteria, determining whether keywords correspond to eligible, ineligible, or trigger categories, and outputting a certification eligibility result. These limitations fall within the certain methods of organizing human activity grouping because they involve following rules or instructions for determining whether a product satisfies certification, regulatory, marketplace, or jurisdiction-specific requirements. The claimed process organizes the interaction between a product, a certification authority or certification framework, and a market/jurisdiction by determining whether the product meets the applicable eligibility rules. The limitations also fall within the mental processes grouping because they encompass concepts that can be performed in the human mind or with pen and paper, such as observing product information, evaluating a specification, identifying product components and materials, determining whether certain words or characteristics are eligible, ineligible, or require review, and making a judgment regarding whether the product is eligible for certification. The recitation of machine learning, visual analysis, OCR, a processor, memory, and outputting an eligibility indication does not remove the claim from the abstract idea and these elements merely automate the collection, classification, evaluation, and presentation of information. The focus of the claim remains on information analysis and decision-making according to certification rules.
Dependent claims 2–7, 9–14, and 16–20 further narrow the abstract idea by reciting additional details of the information analysis and output process. For example, the dependent claims recite extracting textual or visual content, determining that an ineligible keyword results in a not-eligible indication, determining that a trigger keyword results in further review, determining that no ineligible or trigger keywords and at least one eligible keyword results in an eligible indication, specifying that the visual may be drawings, photographs, images, schematics, plans, or blueprints, and performing OCR on the visual. These limitations further describe data gathering, classification, review, and output of certification eligibility, and therefore remain within the mental-process and certain-methods-of-organizing-human-activity groupings.
Independent claim(s) 8 and 15 recite/describe nearly identical steps (and therefore also recite limitations that fall within this subject matter grouping of abstract ideas), and this/these claim(s) is/are therefore determined to recite an abstract idea under the same analysis.
As such, the Examiner concludes that claim 1 recites an abstract idea (Step 2A – Prong One: YES).
Step 2A - Prong Two: In prong two of step 2A, an evaluation is made whether a claim recites any additional element, or combination of additional elements, that integrate the exception into a practical application of that exception. An “addition element” is an element that is recited in the claim in addition to (beyond) the judicial exception (i.e., an element/limitation that sets forth an abstract idea is not an additional element). The phrase “integration into a practical application” is defined as requiring an additional element or a combination of additional elements in the claim to apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception, such that it is more than a drafting effort designed to monopolize the exception.
The requirement to execute the claimed steps/functions using a computer processor, memory, machine learning, OCR techniques, and automation using computer processor, etc. (Claim 1, 8, and 15) is/are equivalent to adding the words “apply it” on a generic computer and/or mere instructions to implement the abstract idea on a generic computer.
Similarly, the limitations of applying a computer processor, memory, a plurality of machine learning models, stored models, a product specification, visual content, a visual analysis technique, OCR or image-derived analysis, etc. (Independent Claim(s) 1, 8, and 15, and dependent claims 5 and 13) are recited at a high level of generality and amount to no more than mere instructions to apply the exception using generic computer components. This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application (see MPEP 2106.05(f)).
Further, the additional limitations beyond the abstract idea identified above, serves merely to generally link the use of the judicial exception to a particular technological environment or field of use. Specifically, it/they serve(s) to limit the application of the abstract idea to computerized environments (e.g., training, storing, accessing, analyzing, outputting, etc. steps performed by a computer processor, memory, machine learning, OCR techniques, and automation using computer processor, etc.). This reasoning was demonstrated in Intellectual Ventures I LLC v. Capital One Bank (Fed. Cir. 2015), where the court determined "an abstract idea does not become nonabstract by limiting the invention to a particular field of use or technological environment, such as the Internet [or] a computer"). This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application (see MPEP 2106.05(h)).
The recited additional element(s) of training models, storing models, accessing a specification, extracting or analyzing product information, identifying a model, determining keywords, classifying keywords, and outputting an eligibility indication constitute insignificant extra-solution activity. These steps amount to data gathering, intermediate data processing, and post-solution presentation of a result. The claimed additional elements merely collect information before the eligibility determination, analyze information during the eligibility determination, and output the result after the determination (Claim(s) 1, 8, and 15), additionally and/or alternatively simply append insignificant extra-solution activity to the judicial exception, (e.g., mere pre-solution activity, such as data gathering, in conjunction with an abstract idea). This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application. (See MPEP 2106.05(g)).
Dependent claims 2–7, 9–14, and 16–20 do not add limitations that integrate the abstract idea into a practical application. Claims 2, 9, and 16 recite extracting textual or visual content and determining keywords, which is data gathering and information analysis. Claims 3, 10, and 17 recite outputting that the product is not eligible when an ineligible keyword is present, which is merely applying a rule and presenting a result. Claims 4, 11, and 18 recite outputting that the specification needs further review when a trigger keyword is present, which is likewise rule-based review and output. Claims 5, 12, and 19 recite outputting that the product is eligible when no ineligible or trigger keyword is present and at least one eligible keyword is present, which is another rule-based eligibility determination. Claims 6, 13, and 20 merely specify the type of visual content, such as drawings, photographs, images, schematics, plans, or blueprints. Claims 7 and 14 recite OCR, which is used only to extract information from the visual for the abstract eligibility determination. In other words, each of the limitations/elements recited in respective dependent claims is/are further part of the abstract idea as identified by the Examiner for each respective dependent claim (i.e., they are part of the abstract idea recited in each respective claim).
The Examiner has therefore determined that the additional elements, or combination of additional elements, do not integrate the abstract idea into a practical application. Accordingly, the claim(s) is/are directed to an abstract idea (Step 2A – Prong two: NO).
Step 2B: In step 2B, the claims are analyzed to determine whether any additional element, or combination of additional elements, is/are sufficient to ensure that the claims amount to significantly more than the judicial exception. This analysis is also termed a search for an "inventive concept." An "inventive concept" is furnished by an element or combination of elements that is recited in the claim in addition to (beyond) the judicial exception, and is sufficient to ensure that the claim as a whole amounts to significantly more than the judicial exception itself. Alice Corp., 134 S. Ct. at 2355, 110 USPQ2d at 1981 (citing Mayo, 566 U.S. at 72-73, 101 USPQ2d at 1966).
As discussed above in “Step 2A – Prong 2”, the identified additional elements in independent claim(s) 1, 8, and 15, and dependent claims 2-7, 9-14, and 16-20 are equivalent to adding the words “apply it” on a generic computer, and/or generally link the use of the judicial exception to a particular technological environment or field of use. Therefore, the claims as a whole do not amount to significantly more than the judicial exception itself.
The recited additional element(s) of receiving or transmitting data, storing and retrieving information in memory, extracting data from a document or image, analyzing the extracted data, and presenting a result. (Claim(s) 1, 8, and 15), additionally and/or alternatively simply append insignificant extra-solution activity to the judicial exception, (e.g., mere pre-solution activity, such as data gathering, in conjunction with an abstract idea) i.e. is similar to “Receiving or transmitting data over a network, e.g., using the Internet to gather data”, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information), “Storing and retrieving information in memory”, Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93; “Presenting offers to potential customers and gathering statistics generated based on the testing about how potential customers responded to the offers; the statistics are then used to calculate an optimized price”, Electronically scanning or extracting data from a physical document, Content Extraction and Transmission, LLC v. Wells Fargo Bank, 776 F.3d 1343, 1348, 113 USPQ2d 1354, 1358 (Fed. Cir. 2014) (optical character recognition), is a well-understood, routine, and conventional function when it is claimed in a merely generic manner (as it is here) (See MPEP 2106.05(d) (II)).
This conclusion is based on a factual determination. Applicant’s own disclosure at paragraph [0025] acknowledges that “server computer 115 may input, into the generated machine learning models, a set of input data (which may be a set of real-world product data) associated with one or more products for which certification eligibility may be desired. In embodiments, the set of input data may include textual and/or visual content describing and/or depicting the one or more products. The machine learning model may output a result which may include an indication of whether the product is eligible for a certification, the product is not eligible for the certification, or the set of input needs further review. A user of the electronic devices 101, 102, 103 (e.g., an entity associated with the product) may review the result(s) or output(s) and make decisions and take actions accordingly. In embodiments, a user may access the result(s) or output(s) directly from the server computer 115” (i.e., conventional nature of receiving and transmitting data/messages over a network). This additional element therefore do not ensure the claim amounts to significantly more than the abstract idea.
Viewing the additional limitations in combination also shows that they fail to ensure the claims amount to significantly more than the abstract idea. When considered as an ordered combination, the additional components of the claims add nothing that is not already present when considered separately, and thus simply append the abstract idea with words equivalent to “apply it” on a generic computer and/or mere instructions to implement the abstract idea on a generic computer or/and append the abstract idea with insignificant extra solution activity associated with the implementation of the judicial exception, (e.g., mere data gathering, post-solution activity) and/or simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception.
The dependent claims 2-7, 9-14, and 16-20 fail to include any additional elements. In other words, each of the limitations/elements recited in respective independent claims is/are further part of the abstract idea as identified by the Examiner for each respective dependent claim (i.e., they are part of the abstract idea recited in each respective claim). Claims 2, 9, and 16 add limitations directed to extracting, by the computer processor from the specification, at least one of textual content or visual content, and determining, from the extracted textual content or visual content using the machine learning model, the set of keywords. These additional limitations merely refine the manner in which product information is obtained, extracted, and analyzed for the underlying certification-eligibility determination. The use of textual content, visual content, a specification, a processor, and a machine learning model is recited at a high level of generality and does not amount to an improvement in computer functionality, machine-learning technology, image-processing technology, or data-extraction technology. Rather, these elements amount to using generic computer tools to gather and organize information for the abstract idea of evaluating product certification eligibility. Claims 3, 10, and 17 add limitations directed to determining that at least one keyword in the set of keywords is an ineligible keyword and, based on that determination, outputting an indication that the product is not eligible for the certification. These limitations merely apply a rule to classified information and present the result of that rule-based determination. The claims do not recite any specific technological improvement in how the keyword is identified, how the eligibility rule is technically implemented, or how the output is generated. Instead, these limitations amount to evaluating product information according to eligibility criteria and presenting a not-eligible result using generic computer functionality. Claims 4, 11, and 18 add limitations directed to determining that at least one keyword in the set of keywords is a trigger keyword and outputting an indication that the specification needs further review for the product to be eligible for the certification. These limitations merely add a conventional review or exception-handling outcome to the abstract certification-eligibility analysis. Determining that information requires further review is a form of rule-based evaluation, judgment, or decision-making that can be performed mentally or by following instructions. The claims do not recite any unconventional technical mechanism for performing the review, resolving the trigger condition, improving review workflows, or improving the operation of the computer itself. Claims 5, 12, and 19 add limitations directed to determining that the set of keywords does not include any ineligible keywords or trigger keywords and includes at least one eligible keyword, and outputting an indication that the product is eligible for the certification. These limitations merely recite the positive outcome of the same rule-based certification analysis. Determining that no disqualifying or review-triggering keyword exists and that at least one eligible keyword exists is an evaluative judgment based on information classification. The claims do not recite a specific technical improvement in keyword detection, machine-learning model operation, certification processing, or computer functionality. Instead, these limitations amount to applying the abstract idea on a generic computer and outputting the result. Claims 6, 13, and 20 add limitations directed to the type of visual included in the specification, namely at least one of drawings, photographs, images, schematics, plans, or blueprints. These limitations merely specify the source or format of information used in the abstract certification-eligibility determination. The use of drawings, photographs, images, schematics, plans, or blueprints does not provide an inventive concept because the claims do not recite any specific improved image-processing technique, drawing-analysis technique, blueprint-analysis technique, or technical improvement in how such visual information is processed. Rather, the limitations amount to using conventional visual information as input data for the abstract analysis. Claims 7 and 14 add limitations directed to performing an optical character recognition (OCR) technique on the visual to determine a set of components and a set of materials associated with the product. These limitations merely use OCR as a generic data-extraction tool to obtain information from visual content for use in the certification-eligibility analysis. The claims do not recite an improved OCR algorithm, an improved computer-vision architecture, an improved technique for recognizing components or materials, or any technical improvement to OCR or image processing itself. OCR, when recited at this level of generality, is a well-understood, routine, and conventional computer function used to extract information from documents or images. Accordingly, the additional limitations of dependent claims 2-7, 9-14, and 16-20 do not add significantly more than the abstract idea. These limitations merely specify additional data sources, extraction techniques, keyword classifications, rule-based outcomes, review outcomes, and visual formats used in the abstract process of determining product certification eligibility. None of these limitations meaningfully limits the abstract idea, improves computer functionality, improves machine-learning technology, improves OCR technology, or solves a specific technical problem in a technological manner.
When viewed as an ordered combination, the additional elements of claims 2-7, 9-14, and 16-20 merely instruct to implement the abstract idea using generic computer components to collect, store, represent, and display information. The claims do not recite any unconventional arrangement of elements, nor do they effect an improvement to computer functionality or another technical field and therefore fail to integrate the abstract concept into a practical application and it is recited at a high level of generality and does not integrate the judicial exception into a practical application.
The Examiner has therefore determined that no additional element, or combination of additional claims elements is/are sufficient to ensure the claim(s) amount to significantly more than the abstract idea identified above (Step 2B: NO).
Therefore, claims 1-20 are not eligible subject matter under 35 USC 101.
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, 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 pre-AIA 35 U.S.C. 103(a) 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.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Pub. 20200364495 (“Stoettinger”) in view of U.S. Pat. 10937033 (“Rodriguez”).
As per claims 1, 8, and 15, Stoettinger discloses, computer-implemented method of using machine learning to determine product certification eligibility, the method comprising (Examiner interprets “product certification eligibility” broadly as a regulatory classification or compliance eligibility determination. A determination that a product is dangerous, non-dangerous, compliant, non-compliant, or belongs to a regulated class is analogous to determining whether the product is eligible for a certification or regulatory approval.) (“assist the process of classifying products into one or more classes of dangerous goods, the example embodiments provide a fully automated system that predictions if a product is a dangerous good or not and estimates its risks. For example, the prediction may be performed using an algorithm which considers whether a product should be classified within any class of dangerous good from among all classes (e.g., explosive, gases, flammable, reactive, toxic, oxidizing, infective, radioactive, corrosive, etc.) set forth by a regulation. In addition, the algorithm can also consider multiple regulations at the same time. The system can not only assist experts in the classification process perform a double-check of already classified products to increase accuracy and perform plausibility checks in related industrial processes”) (0016-0019, 0042-0046, 0051-0053):
training, by a computer processor, a plurality of machine learning models using a set of training data associated with a set of products (Examiner interprets Stoettinger’s predictive algorithm as the claimed machine-learning model. The plurality of models may be interpreted as plural classification outputs, plural regulation-specific classifiers, or an obvious implementation of separate models for different product categories, regulations, or jurisdictions.) (“FIG. 3B illustrates an example of the predictive algorithm 320 for performing the prediction in FIG. 3A, in accordance with an example embodiment. In the example of FIG. 3B, the predictive algorithm 320 is a deep learning neural network, however embodiments are not limited thereto. In the example of FIG. 3B, the deep learning neural network includes an input layer and an output layer as well as various additional hidden layers in between the input layer and the output layer such as one or more dropout layers, one or more pooling layers, one or more gated recurrent unit (GRU) layers, one or more concatenation layers, a predication layer (also referred to as a sigmoid layer), and the like … The attributes of the product include physical properties, chemical properties, etc. The algorithm may apply machine learning techniques such as dropout layer, convolution layer, etc. to identify important segments of text that effect classification (i.e., alphanumeric segments of text and numbers that impacts the labeling, etc.) The predictive algorithm 320 learns over a number of iterations and through error minimization and optimization it can determine which segment(s) of text is important to get the labels correct. One or more GRU layers may capture the text segments and a concatenation layer may combine the segments into a sequence which is passed to a prediction layer (sigmoid layer). We are making a prediction on whether the good belongs in each of the different classes. The different predictions for each of the classes of each of the regulations are generated using the Sigmoid function at the end. It will take as input all of the vectorized information and all the convolution information”) (0037-0040),
the set of training data comprising textual content and visual content corresponding to the set of products (Examiner notes that underlined limitation is disclosed by another prior art. Examiner interprets the product attributes and historical classified-product text as textual training data associated with products.) (“system may store product attributes (chemical composition, characteristics, descriptions, etc.) of each product that is to be classified. The system can receive an identification of a product and perform the classification by retrieving the product attributes and converting them into a single string value (e.g., one long string value). Here, the system may retrieve the alphanumeric descriptions/values of the individual attributes and concatenate the descriptions into one long sequence which can be input into a text-based classification algorithm (e.g., machine learning algorithm, etc.) In some embodiments, the machine learning algorithm is a deep learning neural network, but embodiments are not limited thereto. The machine learning algorithm can perform a classification of the product based on the text included in the single string value. Here, the machine learning algorithm can provide a probability, a yes/no answer, etc., of whether the product should be classified within each of a plurality of different classes of dangerous goods, for a plurality of different regulations.”) (0017, 0033-0039),
wherein the visual content comprises, for each product of the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories (Examiner notes that the underlined limitation is disclosed by another prior art. Examiner interprets Stoettinger’s dangerous-goods classes/regulatory risks) (“The dangerous goods classification task needs to be performed independently for different regulations such as ADR or CFR—see appendix for details. This independent classification task may be performed simultaneously by one model predicting multiple dangerous goods classes for an input. Every regulation has a precise definition when a product must be handled as a dangerous good. Additionally, they define main risks for each product out of nine main risk classes (see appendix) and up to two subsidiary risks … predictive algorithm 320 can be used to solve this text classification problem as a multi-label problem where every risk (including subdivisions) per regulation is one label of the classifier, leading to a total number of dozens of labels for the training task (label encoding). The predictive algorithm 320 may be trained from historical text of already classified products thereby providing a corpus of learning for the predictive algorithm 320. The predictive algorithm may receive the input string 310 which includes the product attributes/chemical properties as one large chunk of text. Within the predictive algorithm 320 may include embedding techniques that convert the whole string information into a vectorized format. Then the predictive algorithm 320 may apply normal machine learning techniques to the text to identify patterns in the text that can be the basis of the classification of the product.”) (0038-0039);
storing the plurality of machine learning models in a memory (Examiner interprets the disclosed data store, memory, and storage architecture as storing the ML model, product data, and instructions used to perform the classification/compliance analysis.) (“Data store 110 may be any query-responsive data source or sources that are or become known, including but not limited to a SQL relational database management system. Data store 110 may include or otherwise be associated with a relational database, a multi-dimensional database, an Extensible Markup Language (XML) document, or any other data storage system that stores structured and/or unstructured data. The data of data store 110 may be distributed among several relational databases, dimensional databases, and/or other data sources. Embodiments are not limited to any number or types of data sources.”) (0024-0027);
accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product (Examiner interprets Examiner interprets the product attributes/input string in Stoettinger and the submitted image/text content) (“the classification may be performed by the same device the receives the input. In this example, the user device 210 display a user interface 212 including a search field 214 that enables a user to input a product identifier such as a product name, a product ID, a serial number, or the like. In response, the product ID may be transmitted to the host platform 220 which retrieves attributes 222 of the product from a data store such as a database, a data file, an external system/resource, or the like …”) (0030-0033),
(ii) identifying a certification (Examiner interprets the identified regulation/classification framework as the claimed certification. A regulatory classification under CFR/ADR/GHS-type rules is treated as a certification or approval framework for whether the product may be handled, shipped, or treated as eligible under the applicable regulation) (“process of classifying products into one or more classes of dangerous goods, the example embodiments provide a fully automated system that predictions if a product is a dangerous good or not and estimates its risks. For example, the prediction may be performed using an algorithm which considers whether a product should be classified within any class of dangerous good from among all classes (e.g., explosive, gases, flammable, reactive, toxic, oxidizing, infective, radioactive, corrosive, etc.) set forth by a regulation. In addition, the algorithm can also consider multiple regulations at the same time. The system can not only assist experts in the classification process perform a double-check of already classified products to increase accuracy and perform plausibility checks in related industrial processes …”) (0016-0019, 0034-0036),
and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction (Examiner interprets different regulatory schemes, such as CFR and ADR, as corresponding to different jurisdictions or geographic markets. The certification/classification is applicable to the jurisdiction because each regulation defines its own requirements for the product class.) (“a product may not have all of the property values and therefore may have a reading of non-applicable. Also provided in the Appendix is an example of some of the regulations and the danger classes that are associated with each regulation. The predictive algorithm of the example embodiments may provide a classification of a product for each class of each regulation that is desired. For example, a regulation may have 9 classes of dangerous goods. In this example, the algorithm can provide 10 predictions including one for each of the 9 types of classes and one for non-dangerous classification. Furthermore, the algorithm can perform the same prediction across multiple regulations at the same time. Each class has different requirements to meet in order to be considered in that danger class (legal reasons). There are slight differences from regulation to regulation. But if something is flammable in the United States, it is likely flammable in Europe, but maybe slight differences in what is flammable.” and “the predictive algorithm can simultaneously predict or otherwise determine whether a good should be classified within any of a plurality of classes among any of a plurality of regulations. Therefore, rather than perform one classification at a time, the algorithm may perform dozens, or even hundreds of predictions at once. Furthermore, in 540 the method may include outputting information about the prediction of whether the object is dangerous for display via a user interface. For example, probabilities of whether a product falls within each of the respective classes of dangerous goods may be output. As another example, a Yes/No answer may be output for each class, or the like.”) (0019-0021 and 0045-0046, 0038-0039);
identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product (Examiner interprets the ML algorithm applied to the product-specific input string as a model corresponding to the product.) (“user inputting a product identifier via a user interface in accordance with an example embodiment. Referring to the example of FIG. 2A, a user device 210 accesses a backend server (host platform 220) for performing a dangerous goods classification. Here, the user device 210 may connect to the host platform 220 via a network connection (e.g., Internet, private network, etc.) However, it should be appreciated that the embodiments are not limited thereto. As another example, the classification may be performed by the same device the receives the input. In this example, the user device 210 display a user interface 212 including a search field 214 that enables a user to input a product identifier such as a product name, a product ID, a serial number, or the like … he predictive algorithm 320 solves a text-classification problem and generate a prediction for multiple labels/classes. In other words, the predictive algorithm 320 does not only predict one whether a product falls into one class of dangerous good, but can simultaneously predict whether the product fits into any of a plurality of classes for a plurality of different regulations. In the example of FIG. 3A, an output 330 is generated based on the results of the predictions by the predictive algorithm 320. Here, the output 330 includes a probability (e.g., a percentage) that a product fits into each of the classes of a plurality of regulations which include different classes of dangerous goods and one class for non-dangerous goods. In this example, the product is predicted to be included in a dangerous class 334 of a first regulation 332, and a dangerous class 338 of a second regulation 336. It should also be appreciated that more than two regulations may be predicated at the same time”) (0030-0035),
and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product (Examiner interprets the regulation-specific or jurisdiction-specific prediction as identifying a model or classification path corresponding to the relevant jurisdiction.) (“dangerous goods classification task needs to be performed independently for different regulations such as ADR or CFR—see appendix for details. This independent classification task may be performed simultaneously by one model predicting multiple dangerous goods classes for an input. Every regulation has a precise definition when a product must be handled as a dangerous good. Additionally, they define main risks for each product out of nine main risk classes (see appendix) and up to two subsidiary risks … predictive algorithm 320 can be used to solve this text classification problem as a multi-label problem where every risk (including subdivisions) per regulation is one label of the classifier, leading to a total number of dozens of labels for the training task (label encoding). The predictive algorithm 320 may be trained from historical text of already classified products thereby providing a corpus of learning for the predictive algorithm 320. The predictive algorithm may receive the input string 310 which includes the product attributes/chemical properties as one large chunk of text.” and “the predictive algorithm can simultaneously predict or otherwise determine whether a good should be classified within any of a plurality of classes among any of a plurality of regulations. Therefore, rather than perform one classification at a time, the algorithm may perform dozens, or even hundreds of predictions at once. Furthermore, in 540 the method may include outputting information about the prediction of whether the object is dangerous for display via a user interface … he predicting may include simultaneously predicting whether the object is included within each of a plurality of different classes of dangerous objects via execution of the text-based machine learning algorithm. In some embodiments, the predicting may include simultaneously predicting whether the object is included within a plurality of different classes of dangerous objects for each of a plurality of different jurisdictions. In some embodiments, the outputting may include outputting a plurality of values corresponding to the plurality of different classes of dangerous objects, respectively, where each value indicates a probability that the object is included within a respective class of dangerous objects.”) (0038-0039 and 0045-0046);
and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to (Examiner interprets the extracted image/text content and product attributes as the claimed specification, including product-related material/composition information) (“FIG. 2B, the system may convert the different values 232 of the attributes 230 into a single input string 240. For example, the system may concatenate the retrieved values 232 into one long string that describes the entire database information of the product as a long sequence of textual information. The system may execute an algorithm that builds a feature string per product that is near English text and human-readable. The conversion from the tabular product definition into a string is a deterministic and rule-based process, which can be performed for any product. This feature string is the basis for subsequent training and inference of the classifier. The fact that the feature strings are legible is important to fulfill an important regulatory constraint: the intrinsic need to be able to explain the reasons for a DG classification to the authorities as is further shown in the example of FIG. 4. The resulting input string 240 includes a long concatenation of the attribute values (string) and is in a format that can be input into a machine learning algorithm such as a deep learning neural network which then predicts the dangerous goods classifications for all regulations at once”) (0033-0040):
determine a set of keywords associated with the product (Examiner interprets Stoettinger’s important text segments/substrings and Rodriguez’s blacklisted words, phrases, strings, and image identifiers as the claimed keywords.) (“The dangerous goods classification task needs to be performed independently for different regulations such as ADR or CFR—see appendix for details. This independent classification task may be performed simultaneously by one model predicting multiple dangerous goods classes for an input. Every regulation has a precise definition when a product must be handled as a dangerous good. Additionally, they define main risks for each product out of nine main risk classes (see appendix) and up to two subsidiary risks … predictive algorithm 320 can be used to solve this text classification problem as a multi-label problem where every risk (including subdivisions) per regulation is one label of the classifier, leading to a total number of dozens of labels for the training task (label encoding). The predictive algorithm 320 may be trained from historical text of already classified products thereby providing a corpus of learning for the predictive algorithm 320. The predictive algorithm may receive the input string 310 which includes the product attributes/chemical properties as one large chunk of text. Within the predictive algorithm 320 may include embedding techniques that convert the whole string information into a vectorized format. Then the predictive algorithm 320 may apply normal machine learning techniques to the text to identify patterns in the text that can be the basis of the classification of the product.”) (0038-0040, 0047), wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword (Examiner notes that the underlined limitation is disclosed by another prior art),
and output an indication of whether the product is eligible for the certification (Examiner interprets the output dangerous/non-dangerous classification, probability, yes/no answer, or warning as the claimed indication of whether the product is eligible.) (“the predictive algorithm 320 solves a text-classification problem and generate a prediction for multiple labels/classes. In other words, the predictive algorithm 320 does not only predict one whether a product falls into one class of dangerous good, but can simultaneously predict whether the product fits into any of a plurality of classes for a plurality of different regulations. In the example of FIG. 3A, an output 330 is generated based on the results of the predictions by the predictive algorithm 320. Here, the output 330 includes a probability (e.g., a percentage) that a product fits into each of the classes of a plurality of regulations which include different classes of dangerous goods and one class for non-dangerous goods. In this example, the product is predicted to be included in a dangerous class 334 of a first regulation 332, and a dangerous class 338 of a second regulation 336. It should also be appreciated that more than two regulations may be predicated at the same time”) (0035-0036, 0045-0046, Figs. 4-5),
wherein the indication is based at least on the type of each keyword of the set of keywords (Examiner interprets the output as being based on whether the identified keyword/image/text is compliant, non-compliant, dangerous, non-dangerous, warning-triggering, or approved under the applicable policy/regulation) (“predictive algorithm 320 can be used to solve this text classification problem as a multi-label problem where every risk (including subdivisions) per regulation is one label of the classifier, leading to a total number of dozens of labels for the training task (label encoding). The predictive algorithm 320 may be trained from historical text of already classified products thereby providing a corpus of learning for the predictive algorithm 320. The predictive algorithm may receive the input string 310 which includes the product attributes/chemical properties as one large chunk of text …” and “FIG. 5, in some embodiments, the method may further include identifying a sub-string within the input string that has a most impact on the prediction of whether the object is a dangerous object, and displaying the sub-string via the user interface. For example, the sub-string may include a chunk of text that impacts the reason for classifying the product as a particular class among the plurality of classes of dangerous goods or non-dangerous. In some embodiments, the sub-string may be identified by dividing the input string in half and determining which half has the most impact, and iteratively repeating the dividing and the determining a plurality of times to identify the sub-string. For example, the dividing and determining may be performed a predetermined number of times, until the sub-string is below a predetermined size, a random amount, and the like.”) (0039-0040 and 0047).
Stoettinger specifically doesn’t disclose, and visual content corresponding to the set of products, wherein the visual content comprises, for each product of the set of products, a training visual that visually depicts that product and performing, by the computer processor, a visual analysis technique on the visual to determine a set of components and a set of materials associated with the product, however Rodriguez discloses, and visual content corresponding to the set of products (Examiner interprets adding visual content and image-based analysis to the product classification/compliance workflow) (“retrieve 128 the image data 116 using the image key 118 provided when the user 102 submitted the content 106. In accordance with at least one embodiment, an OCR engine 130 may execute optical character recognition 132 on the image data 116 retrieved at 128. In embodiments, the OCR engine 130 may be configured to identify text, text strings, text phrases, and images included in the image data 116 or submitted content 106 using recognition algorithms such as optical character recognition, image recognition, or object recognition. In workflow 100 the OCR engine 130 may store the text from the image 134 in the extracted data store 136. In accordance with at least one embodiment, the service provider computers 120 may pre-moderate the data at 138 by comparing the extracted data 136 to one or more blacklists of text and/or images according to compliance policies maintained by the service provider computers 120 as well as determine whether any promotional messages violate the compliance policies using the web page data 124. In use cases where the service provider computers 120 identify non-compliant content,”) (Col. 4 Ln. 15-37, Cols. 5-6, Figs. 1 and 4),
wherein the visual content comprises, for each product of the set of products, a training visual that visually depicts that product (Examiner interprets the submitted item images as visuals depicting products. This limitation is stronger for “visual content” and “visual analysis” than for the exact phrase “training visual.” If challenged, add a labeled image-training reference) (“retrieve 128 the image data 116 using the image key 118 provided when the user 102 submitted the content 106. In accordance with at least one embodiment, an OCR engine 130 may execute optical character recognition 132 on the image data 116 retrieved at 128. In embodiments, the OCR engine 130 may be configured to identify text, text strings, text phrases, and images included in the image data 116 or submitted content 106 using recognition algorithms such as optical character recognition, image recognition, or object recognition. In workflow 100 the OCR engine 130 may store the text from the image 134 in the extracted data store 136. In accordance with at least one embodiment, the service provider computers 120 may pre-moderate the data at 138 by comparing the extracted data 136 to one or more blacklists of text and/or images according to compliance policies maintained by the service provider computers 120 as well as determine whether any promotional messages violate the compliance policies using the web page data 124. In use cases where the service provider computers 120 identify non-compliant content,”) (Col. 4 Ln. 15-37, Cols. 5-6);
performing, by the computer processor, a visual analysis technique on the visual to determine a set of components and a set of materials associated with the product (Examiner interprets OCR, image recognition, and object recognition as the claimed visual analysis technique.) (“service provider computers (service provider computers 120 and 614) utilizing at least the pre-moderation module 630 depicted in FIGS. 1 and 6 may perform the processes 400 and 500 of FIGS. 4 and 5. In FIG. 4, the process 400 may include receiving data for a content submission for a web store associated with an electronic marketplace prior to the content submission being incorporated into the web store at 402. For example, a user that utilizes a web store associated with an electronic marketplace may wish to provide an update to their web store, perhaps to sell new products for example. The user may interact with a user interface and provide new content such as images or additional text that they wish to incorporate into their web store. Conventionally a user may interact with a submit content or submit for publishing button or element of the user interface to initiate a process where manual moderation may identify whether the newly submitted content (images or text) includes non-compliant content. The process 400 may include extracting text and images in the data based at least in part on a recognition algorithm at 404. In embodiments, the service provider computers implementing the pre-moderation feature may utilize optical character recognition algorithm, an image recognition algorithm or an object detection algorithm to identify the text and images included in the content”) (Col. 6 Ln. 13-35, Cols. 4-6);
wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword (Examiner interprets “ineligible keyword” as blacklisted/non-compliant or dangerous-class text; “trigger keyword” as text/image content producing a warning, threshold issue, or further moderation review; and “eligible keyword” as compliant/non-dangerous text supporting approval or non-dangerous classification) (“compliance policies may be utilized to determine if an identified text object or image object includes non-compliant content. For example, compliance policies may include policies for identifying whether submitted text includes incorrect grammar, punctuation, spelling, font size, or capitalization errors. Compliance polices may identify whether an image or text includes promotional messages that can't be supported by an associated web store, whether an image is cropped incorrectly, or utilizes a resolution that is inappropriate for the web site, whether images are obscured by other objects, or whether hyperlinks included in the content are broken. In accordance with at least one embodiment, the pre-moderation feature may generate one or more warnings or recommendations and update a user interface or web browser to inform the user about the non-compliant content as well as give the user an opportunity to modify previously submitted text or images and correct the non-compliant content. In embodiments, the user interface or web browser may visually highlight or otherwise indicate portions of the web site, web store, or content that corresponds to the non-compliant content included in the warnings or recommendations.”) (Col. 2 Ln. 21-43).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention to training a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories, storing the plurality of machine learning models in a memory, accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction, identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product, and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to: determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords, as disclosed by Stoettinger, and visual content corresponding to the set of products, wherein the visual content comprises, for each product of the set of products, a training visual that visually depicts that product and performing a visual analysis technique on the visual to determine a set of components and a set of materials associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, as taught by Rodriguez for the purpose to analyze not only textual product attributes but also submitted product images and image-derived text, thereby increasing the accuracy and completeness of the product compliance/classification determination thus to improve product classification by allowing visual product information to be considered along with product attributes and regulatory data.
As per claims 2, 9, and 16, Stoettinger discloses, analyzing, using the machine learning model, the specification comprises: and determining, by the computer processor from the at least one of the set of textual content or the set of visual content using the machine learning model, the set of keywords (Examiner interprets converts product attributes into input text and uses ML to identify patterns/segments affecting classification.) (“system may convert the different values 232 of the attributes 230 into a single input string 240. For example, the system may concatenate the retrieved values 232 into one long string that describes the entire database information of the product as a long sequence of textual information. The system may execute an algorithm that builds a feature string per product that is near English text and human-readable. The conversion from the tabular product definition into a string is a deterministic and rule-based process, which can be performed for any product. This feature string is the basis for subsequent training and inference of the classifier.”) (0033-0040, 0047, 0052)
Stoettinger specifically doesn’t disclose, extracting, by the computer processor from the specification, at least one of a set of textual content or a set of visual content, however Rodriguez discloses, extracting, by the computer processor from the specification, at least one of a set of textual content or a set of visual content (Examiner interprets “ineligible keyword” as blacklisted/non-compliant or dangerous-class text; “trigger keyword” as text/image content producing a warning, threshold issue, or further moderation review; and “eligible keyword” as compliant/non-dangerous text supporting approval or non-dangerous classification) (“The pre-moderation feature described herein can identify non-compliant content and warn a user submitting the content such that they can correct the content prior to review by a moderation service. In accordance with at least one embodiment, the pre-moderation feature can extract text and images in content that is submitted by a user for incorporation to a web site or web store using recognition algorithms such as optical character recognition (OCR), image recognition, or object recognition. The pre-moderation feature may generate a score for each extracted text object or image object that corresponds to a confidence in identification of the text object or image object by the recognition algorithms. In embodiments, service provider computers implementing the pre-moderation feature may use one or more thresholds to determine if the confidence score exceeds the threshold and that the item has been correctly identified. In cases where a text object or image object has been properly identified, the object can be analyzed to determine if it includes non-compliant content.”) (Col. 2 Ln. 2-43, Col. 3 Ln. 20-30, Col. 6 Ln. 25-65).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention to training a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories, storing the plurality of machine learning models in a memory, accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction, identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product, and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to: determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords, as disclosed by Stoettinger, extracting, by the computer processor from the specification, at least one of a set of textual content or a set of visual content, as taught by Rodriguez for the purpose to analyze not only textual product attributes but also submitted product images and image-derived text, thereby increasing the accuracy and completeness of the product compliance/classification determination thus to improve product classification by allowing visual product information to be considered along with product attributes and regulatory data.
As per claims 3, 10, and 17, Stoettinger discloses, and wherein outputting the indication of whether the product is eligible for the certification comprises: outputting the indication that the product is not eligible for the certification (Examiner interprets converts product attributes into input text and uses ML to identify patterns/segments affecting classification.) (“system may convert the different values 232 of the attributes 230 into a single input string 240. For example, the system may concatenate the retrieved values 232 into one long string that describes the entire database information of the product as a long sequence of textual information. The system may execute an algorithm that builds a feature string per product that is near English text and human-readable. The conversion from the tabular product definition into a string is a deterministic and rule-based process, which can be performed for any product. This feature string is the basis for subsequent training and inference of the classifier.”) (0033-0040, 0044-0047, 0052)
Stoettinger specifically doesn’t disclose, determining that at least one keyword in the set of keywords is the ineligible keyword, however Rodriguez discloses, determining that at least one keyword in the set of keywords is the ineligible keyword (Examiner interprets “ineligible keyword” as blacklisted/non-compliant or dangerous-class text; “trigger keyword” as text/image content producing a warning, threshold issue, or further moderation review; and “eligible keyword” as compliant/non-dangerous text supporting approval or non-dangerous classification) (“service provider computers (service provider computers 120 and 614) utilizing at least the pre-moderation module 630 depicted in FIGS. 1 and 6 may perform the processes 400 and 500 of FIGS. 4 and 5. In FIG. 4, the process 400 may include receiving data for a content submission for a web store associated with an electronic marketplace prior to the content submission being incorporated into the web store at 402. For example, a user that utilizes a web store associated with an electronic marketplace may wish to provide an update to their web store, perhaps to sell new products for example. The user may interact with a user interface and provide new content such as images or additional text that they wish to incorporate into their web store. Conventionally a user may interact with a submit content or submit for publishing button or element of the user interface to initiate a process where manual moderation may identify whether the newly submitted content (images or text) includes non-compliant content. The process 400 may include extracting text and images in the data based at least in part on a recognition algorithm at 404. In embodiments, the service provider computers implementing the pre-moderation feature may utilize optical character recognition algorithm, an image recognition algorithm or an object detection algorithm to identify the text and images included in the content”) (Col. 6 Ln. 13-35, Cols. 4-6).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention to training a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories, storing the plurality of machine learning models in a memory, accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction, identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product, and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to: determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords, as disclosed by Stoettinger, determining that at least one keyword in the set of keywords is the ineligible keyword, as taught by Rodriguez for the purpose to analyze not only textual product attributes but also submitted product images and image-derived text, thereby increasing the accuracy and completeness of the product compliance/classification determination thus to improve product classification by allowing visual product information to be considered along with product attributes and regulatory data.
As per claims 4, 11, and 18, Stoettinger discloses, and wherein outputting the indication of whether the product is eligible for the certification comprises: outputting the indication that the specification needs further review for the product to be eligible for the certification (Examiner interprets converts product attributes into input text and uses ML to identify patterns/segments affecting classification.) (“system may convert the different values 232 of the attributes 230 into a single input string 240. For example, the system may concatenate the retrieved values 232 into one long string that describes the entire database information of the product as a long sequence of textual information. The system may execute an algorithm that builds a feature string per product that is near English text and human-readable. The conversion from the tabular product definition into a string is a deterministic and rule-based process, which can be performed for any product. This feature string is the basis for subsequent training and inference of the classifier.”) (0033-0040, 0044-0047, 0052)
Stoettinger specifically doesn’t disclose, determining that at least one keyword in the set of keywords is the trigger keyword, however Rodriguez discloses, determining that at least one keyword in the set of keywords is the trigger keyword (Examiner interprets “trigger keyword” as blacklisted/non-compliant or dangerous-class text; “trigger keyword” as text/image content producing a warning, threshold issue, or further moderation review; and “eligible keyword” as compliant/non-dangerous text supporting approval or non-dangerous classification) (“service provider computers (service provider computers 120 and 614) utilizing at least the pre-moderation module 630 depicted in FIGS. 1 and 6 may perform the processes 400 and 500 of FIGS. 4 and 5. In FIG. 4, the process 400 may include receiving data for a content submission for a web store associated with an electronic marketplace prior to the content submission being incorporated into the web store at 402. For example, a user that utilizes a web store associated with an electronic marketplace may wish to provide an update to their web store, perhaps to sell new products for example. The user may interact with a user interface and provide new content such as images or additional text that they wish to incorporate into their web store. Conventionally a user may interact with a submit content or submit for publishing button or element of the user interface to initiate a process where manual moderation may identify whether the newly submitted content (images or text) includes non-compliant content. The process 400 may include extracting text and images in the data based at least in part on a recognition algorithm at 404. In embodiments, the service provider computers implementing the pre-moderation feature may utilize optical character recognition algorithm, an image recognition algorithm or an object detection algorithm to identify the text and images included in the content”) (Col. 6 Ln. 13-55, Col. 7 ln. 1-35, Cols. 4-6).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention to training a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories, storing the plurality of machine learning models in a memory, accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction, identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product, and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to: determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords, as disclosed by Stoettinger, determining that at least one keyword in the set of keywords is the trigger keyword, as taught by Rodriguez for the purpose to analyze not only textual product attributes but also submitted product images and image-derived text, thereby increasing the accuracy and completeness of the product compliance/classification determination thus to improve product classification by allowing visual product information to be considered along with product attributes and regulatory data.
As per claims 5, 12, and 19, Stoettinger discloses, and wherein outputting the indication of whether the product is eligible for the certification comprises: outputting the indication that the product is eligible for the certification (Examiner interprets Stoettinger outputs probabilities, yes/no answers, colored/visual identifiers, class descriptions, and rules/regulations associated with the dangerous-goods classification. Stoettinger also outputs a non-dangerous classification) (“system may convert the different values 232 of the attributes 230 into a single input string 240. For example, the system may concatenate the retrieved values 232 into one long string that describes the entire database information of the product as a long sequence of textual information. The system may execute an algorithm that builds a feature string per product that is near English text and human-readable. The conversion from the tabular product definition into a string is a deterministic and rule-based process, which can be performed for any product. This feature string is the basis for subsequent training and inference of the classifier.”) (0035-0036, 0033-0040, 0044-0047, 0052).
Stoettinger specifically doesn’t disclose, determining that the set of keywords (i) does not include any ineligible keywords or trigger keywords, and (ii) includes at least one eligible keyword, however Rodriguez discloses, determining that the set of keywords (i) does not include any ineligible keywords or trigger keywords, and (ii) includes at least one eligible keyword (Examiner interprets Rodriguez discloses compliance policies and non-compliant determinations; content not matching blacklists/non-compliant policies may be treated as compliant. Rodriguez also maintains blacklists of words/images and applies compliance policies.) (“The system can generate and communicate warnings or recommendations to users such that the users have an opportunity to correct the non-compliant content and thus reduce the overall moderation turnover for their web site or web store. Further, as the pre-moderation feature learns from submitted content, updated blacklist libraries, and user feedback, users can bypass review by a human operated moderation service such that compliant content can be incorporated into a web site or web store with in a manner of minutes as opposed to days or weeks.”) (Col. 3 Ln. 25-35, Col. 7 ln. 1-35, Cols. 4-6).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention to training a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories, storing the plurality of machine learning models in a memory, accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction, identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product, and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to: determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords, as disclosed by Stoettinger, determining that the set of keywords (i) does not include any ineligible keywords or trigger keywords, and (ii) includes at least one eligible keyword, as taught by Rodriguez for the purpose to analyze not only textual product attributes but also submitted product images and image-derived text, thereby increasing the accuracy and completeness of the product compliance/classification determination thus to improve product classification by allowing visual product information to be considered along with product attributes and regulatory data.
As per claims 6, 13, and 20, Stoettinger discloses, wherein accessing the specification comprising the visual that visually depicts the product comprises: accessing, by the computer processor, the specification comprising at least one of: a set of drawings, a set of photographs, a set of images, a set of schematics, a set of plans, or a set of blueprints (Examiner interprets the receiving a product identifier and retrieving product attributes from a data store, including chemical composition, characteristics, descriptions, physical/chemical properties, and regulatory information and also claim recites “at least one of” drawings, photographs, images, schematics, plans, or blueprints) (“product attributes (chemical composition, characteristics, descriptions, etc.) of each product that is to be classified. The system can receive an identification of a product and perform the classification by retrieving the product attributes and converting them into a single string value (e.g., one long string value). Here, the system may retrieve the alphanumeric descriptions/values of the individual attributes and concatenate the descriptions into one long sequence which can be input into a text-based classification algorithm (e.g., machine learning algorithm, etc.) In some embodiments, the machine learning algorithm is a deep learning neural network, but embodiments are not limited thereto. The machine learning algorithm can perform a classification of the product based on the text included in the single string value. Here, the machine learning algorithm can provide a probability, a yes/no answer, etc., of whether the product should be classified within each of a plurality of different classes of dangerous goods, for a plurality of different regulations.”) (0017, 0030-0033, 0044-0047, 0052).
Examiner notes that Rodriguez also discloses receiving submitted content that includes images, image data, and text/images extracted from the content. Rodriguez’s Fig. 1 shows image data 116, OCR engine 130, and text-from-image 134; the Abstract states that text and images are extracted using recognition algorithms.
As per claims 7 and 14, Stoettinger discloses, to determine a set of components and a set of materials associated with the product (Examiner interprets that Stoettinger discloses product attributes including chemical composition, characteristics, descriptions, physical properties, chemical properties, safety, toxicology, and regulations. The ML model uses these attributes to classify the product.) (“system may convert the different values 232 of the attributes 230 into a single input string 240. For example, the system may concatenate the retrieved values 232 into one long string that describes the entire database information of the product as a long sequence of textual information. The system may execute an algorithm that builds a feature string per product that is near English text and human-readable. The conversion from the tabular product definition into a string is a deterministic and rule-based process, which can be performed for any product. This feature string is the basis for subsequent training and inference of the classifier.”) (0017, 0033-0040, 0044-0047, 0052)
Stoettinger specifically doesn’t disclose, wherein performing the visual analysis technique on the visual comprises: performing, by the computer processor, an optical character recognition (OCR) technique on the visual, however Rodriguez discloses, wherein performing the visual analysis technique on the visual comprises: performing, by the computer processor, an optical character recognition (OCR) technique on the visual (Examiner interprets OCR/image/object recognition as the claimed visual analysis technique.) (“retrieve 128 the image data 116 using the image key 118 provided when the user 102 submitted the content 106. In accordance with at least one embodiment, an OCR engine 130 may execute optical character recognition 132 on the image data 116 retrieved at 128. In embodiments, the OCR engine 130 may be configured to identify text, text strings, text phrases, and images included in the image data 116 or submitted content 106 using recognition algorithms such as optical character recognition, image recognition, or object recognition. In workflow 100 the OCR engine 130 may store the text from the image 134 in the extracted data store 136. In accordance with at least one embodiment, the service provider computers 120 may pre-moderate the data at 138 by comparing the extracted data 136 to one or more blacklists of text and/or images according to compliance policies maintained by the service provider computers 120 as well as determine whether any promotional messages violate the compliance policies using the web page data 124. In use cases where the service provider computers 120 identify non-compliant content,”) (Col. 4 Ln. 15-37, Cols. 5-6, Figs. 1 and 4).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention to training a plurality of machine learning models using a set of training data associated with a set of products, the set of training data comprising textual content and visual content corresponding to the set of products, a training visual that visually depicts that product, and wherein the training visual comprises a label for a category of a plurality of categories, storing the plurality of machine learning models in a memory, accessing, by the computer processor, a specification (i) comprising a visual that visually depicts a product, (ii) identifying a certification, and (iii) indicating a geographic area or jurisdiction corresponding to an envisioned market for the product, wherein the certification is applicable to the geographic area or jurisdiction, identifying, by the computer processor, a machine learning model of the plurality of machine learning models that corresponds to (i) the product, and (ii) the geographic area or jurisdiction corresponding to the envisioned market for the product, and analyzing, by the computer processor using the machine learning model, the specification, including the set of components and the set of materials to: determine a set of keywords associated with the product, wherein each keyword of the set of keywords has a type that is one of an ineligible keyword, a trigger keyword, or an eligible keyword, and output an indication of whether the product is eligible for the certification, wherein the indication is based at least on the type of each keyword of the set of keywords, as disclosed by Stoettinger, wherein performing the visual analysis technique on the visual comprises: performing, by the computer processor, an optical character recognition (OCR) technique on the visual, as taught by Rodriguez for the purpose to analyze not only textual product attributes but also submitted product images and image-derived text, thereby increasing the accuracy and completeness of the product compliance/classification determination thus to improve product classification by allowing visual product information to be considered along with product attributes and regulatory data.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure. The following references have been cited to further show the state of the art with respect to analyze to identify one or both of classification values and categories to generate training data that is added to the training set and keywords that are unacceptable for publishing via the content publishing platform.
U.S. Pub. No. 20150120363 (“Yoshinaga”)
Yoshinaga discloses, method for obtaining product related information. Information related to a product obtained from a plurality of different sources is transformed into processed product data with a plurality of levels. Callouts and contexts are identified in the processed product data. A product-to-chemical continuum is generated by creating callout-context pathway segments between the plurality of levels of the processed product data based on the callouts and contexts identified. A query request for product information is transformed into a set of context search parameters, which is used to traverse the product-to-chemical continuum through the callout-context pathway segments that span the plurality of levels. The product information that matches the set of context search parameters is extracted from the product-to-chemical continuum. The callout-context pathway segments reduce processing resources and time needed to obtain the product information.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GAUTAM UBALE whose telephone number is (571)272-9861. The examiner can normally be reached Mon-Fri. 7:00 AM- 6:30 PM PST.
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, Marissa Thein can be reached on (571) 272-6764. 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.
/GAUTAM UBALE/Primary Examiner, Art Unit 3689
Prosecution Insights