DETAILED ACTION
This is a Final Office action is in response to communications filed on May 12th, 2026. Claim 1 and 5 is/are amended and new claim 9 is/are added.
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 .
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 1-9 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 1-4 is/are drawn to method (i.e., a process), and claims 5-8 is/are drawn to system (i.e., a manufacture), and claims 9 is/are drawn to computer readable medium (i.e., a manufacture). (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.
Claim 1: A method for identifying a match with a product, comprising
extracting, by a processor, one or more attributes relating to a product from data indicated in a request for identifying the product, each of the one or more attributes describing the product, wherein the data comprises an image of the product and extracting the one or more attributes comprises identifying the one or more attributes from the image based on a visual question and answering (VQA) paradigm wherein the VQA paradigm comprises generating (i) a question embedding, (ii) a plurality of text embeddings derived from optical character recognition (OCR) on the image, and (iii) a plurality of image embeddings generated from image patches and processing the embeddings using a neural network;
adapting, by the processor, the one or more attributes based on historical data relating to the product into a same format as that of one or more entries in a database, the database comprising one or more entries for each of a plurality of products;
identifying a match with the product from the plurality of products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the plurality of products;
selecting one or more products from the plurality of products in the database based on a rule-based comparison of at least one of the one or more attributes with each corresponding entry in the database;
identifying the match with the product from the selected one or more products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the selected one or more products;
calculating a score for each of the selected one or more products based on the comparison;
identifying the match with the product when the score for the match is a highest score of the calculated scores and greater than a threshold value;
verifying whether each of the selected one or more products excluding the identified match having a score greater than the threshold value is also a match with the product;
and retraining the processor based on the verification.
(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 claim 1 is/are directed to the abstract idea of organizing, analyzing, and evaluating information to identify a matching product, which falls within the category of data analysis and information processing. Specifically, the claim recites collecting product-related data, extracting descriptive attributes from an image, normalizing the attributes using historical information, comparing the attributes against database entries, applying rule-based filtering, scoring and ranking candidate products, selecting a highest-scoring match based on a threshold, verifying alternative candidates, and updating the model based on the verification. These steps collectively describe a process of analyzing information, applying rules to classify and rank data, and making a selection based on evaluation criteria, which are fundamental practices long used in commerce and decision-making. The use of a visual question and answering (VQA) paradigm and retraining does not change the focus of the claim, as they are merely techniques for extracting and refining information and do not alter the underlying abstract idea of comparing product attributes to identify a best match. Thus the claimed subject matter is directed to an abstract ideas falling within the judicial exception category of “mental processes” and “certain methods of organizing human activity”. From applicant’s specification, the claimed invention is implemented to “Figs. 8A and 8B form a schematic block diagram of a general purpose computer system upon which the transaction processing server of Fig. 1 can be practiced.[0016] Fig. 8C is a schematic block diagram of a general purpose computer system upon which the product matching server of Fig. 2 can be practiced” (see 0015-17 of instant specification). The steps under its broadest reasonable interpretation specifically directed to an abstract idea of analyzing information, applying rules, and evaluating data to make a selection of product, which is an instance of certain methods of organizing human activity and mental processes. The Examiner notes that although the claim limitations are summarized, the analysis regarding subject matter eligibility considers the entirety of the claim and all of the claim elements individually, as a whole, and in ordered combination.
And the dependent claims 2-4, and 6-8 recites an abstract idea, directed to analyzing previously extracted product attributes to identify “key” attributes, selecting products by comparing those key attributes against database entries, and normalizing attributes into common formats such as units of measurement, quantities, languages, or item names to facilitate comparison. These steps describe organizing, filtering, and evaluating information using rules and comparisons, which are fundamental data analysis activities that can be performed mentally or with pen and paper and have long been used in commercial product matching and decision-making. Identifying key attributes involves judgment and prioritization of information, selecting products based on attribute comparisons involves applying rules to classify data, and adapting attributes into standardized formats involves routine normalization of information. The claims merely automate these abstract practices using a processor and a database and do not recite any improvement to computer technology or any specialized computing components. As such, the claims are directed to abstract ideas rather than a technological improvement. As such, the claims are directed to an abstract idea involving certain methods of organizing human activity and mental processes, which falls within a judicial exception under 35 U.S.C. §101.
Independent claim(s) 5 and 9 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 claims 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 processor, memory, etc. (Claims 1, 5, and 9) 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 using a processor, memory, etc. (Claims 1, 5, and 9, and dependent claims 2-4 and 6-8) 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., extract, adapt, identify, select, calculate, verify, retrain, etc. steps performed by a processor, memory, 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 receiving product-related data in a request, obtaining an image of the product, accessing historical information and database entries, and generating question, OCR derived text, and image patch embeddings merely gather, retrieve, encode, and prepare the information used to perform the abstract product-matching process. The embeddings merely place the collected question, text, and image information into forms suitable for subsequent analysis by the neural network. To the extent these receiving, obtaining, OCR processing, database accessing, and embedding generation limitations are considered additional elements rather than part of the judicial exception, they merely append insignificant pre-solution activity to the abstract idea, including collecting product information and converting that information into representations used for comparison, classification, scoring, and selection. Similarly, retraining the processor based on the verification merely uses the results of the claimed evaluation to refine the same product-matching process for future iterations and does not apply the identified match to produce a separate technological result. The received request, image, historical data, database entries, and generated embeddings merely supply the information and evaluation criteria used to identify the matching product. The retraining limitation merely updates the information analysis process based on the outcome of that evaluation. These limitations do not impose a meaningful limit on practicing the abstract idea and therefore do not integrate the judicial exception into a practical application. 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 2-4 and 6-8 fail to include any additional elements. 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 Claims 1, 5, and 9, and dependent claims 2-4 and 6-8 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 product-related data in a request, obtaining an image of the product, accessing historical information and database entries, and generating question, OCR derived text, and image patch embeddings merely gather, retrieve, encode, and prepare the information used to perform the abstract product-matching process. The embeddings merely place the collected question, text, and image information into forms suitable for subsequent analysis by the neural network (Independent Claims 1, 5, and 9), 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. extracting attributes from an image using a visual question and answering (VQA) paradigm, comparing extracted attributes with database entries, selecting products based on rule-based comparisons, scoring and ranking candidate products, and selecting a highest-scoring product based on a threshold constitute insignificant extra-solution activity that does not meaningfully limit the judicial exception. These steps merely describe conventional data gathering, analysis, comparison, and post-solution evaluation activities performed to implement the underlying abstract idea of identifying a matching product based on attribute comparison. These steps amount to data gathering, data transmission, and post-solution activity, which 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”, OIP Technologies, 788 F.3d at 1363, 115 USPQ2d at 1092-93, Determining an estimated outcome and setting a price, OIP Techs., 788 F.3d at 1362-63, 115 USPQ2d at 1092-93, 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 [0070] acknowledges that “product matching server 140 may also comprise a matching module 266 that is configured for identifying a match with the product from the plurality of products, the identification based on comparing the one or more attributes with each corresponding entry of the plurality of products in the reference database 150. Identifying the match with the product from the plurality of products may further comprise selecting one or more products from the plurality of products in the reference database 150 based on a rule- based comparison of at least one of the one or more attributes with each corresponding entry in the list, and identifying the match with the product from the selected one or more products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the selected one or more products”. 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 2-4 and 6-8 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 2-4 and 6-8 merely recite identifying one or more “key” attributes from previously extracted attributes and selecting products by comparing those key attributes with database entries. These limitations merely refine the abstract idea by prioritizing, filtering, and comparing information, which are well-understood, routine, and conventional data analysis techniques. Further the claims recites adapting attributes into standardized formats, such as a unit of measurement, quantity, language, or item name, to match database entries. Collectively, these dependent claims constitute well-understood, routine, and conventional activities performed by generic computer components 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 1-9 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 of this title, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
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 factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
Determining the scope and contents of the prior art.
Ascertaining the differences between the prior art and the claims at issue.
Resolving the level of ordinary skill in the pertinent art.
Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-9 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Pat. 11282124 (“Yoon”) in view U.S. Pub. 20170287038 (“Krasadakis”) in further view of NPL “LaTr - Layout-Aware Transformer for Scene-Text Visual Question Answer” (“Biten”).
As per claims 1, 5, and 9, Yoon discloses, extracting, by a processor, one or more attributes relating to a product from data indicated in a request for identifying the product, each of the one or more attributes describing the product (Examiner interprets that Yoon discloses a method performed by an attribute-analysis system for identifying or recommending products in response to user interactions and requests. Yoon’s attribute-analysis system analyzes information associated with items in an electronic catalog, determines descriptive attributes, and identifies or recommends catalog items based on those attributes) (“FIG. 1A illustrates a block diagram of an attribute analysis system 100 in communication with a user device 150. The attribute analysis system 100 may be a system of one or more computers, one or more virtual machines executing on a system of one or more computers, and so on. As described herein, the attribute analysis system 100 may determine item recommendations 122 based on analyzing user interactions 102 of users with respect to an electronic catalog 112. In the illustrated example, the electronic catalog 112 is represented as information included in a database or storage system. As described above, the electronic catalog 112 may include information associated with thousands, millions or billions of items. These items may be available via an electronic store, and may be offered by different third-party entities”) (Col. 5 Ln. 40-46, Col. 11, Ln. 29-37, Fig. 1A), wherein the data comprises an image of the product (Examiner interprets user interfaces display product images and operate on visually presented items in the electronic catalog. Product images constitute part of the data used by the system when responding to user requests. Thus, the data from which attributes are extracted reasonably includes images of the products i.e. Yoon further discloses an electronic catalog containing information associated with thousands, millions, or billions of items and generating product recommendations in response to a search query or request) (“FIG. 1B is an example user interface 160 illustrating recommended items 162 according to the techniques described herein. The example user interface 160 may represent an embodiment of the user interface 152 in which the user is searching the electronic catalog 112 for hats. For example, the user may optionally provide a search query or request to find hats. As another example, the user may select which are commonly viewed, or selected for checkout, by other users”) (Col. 11 Ln. 29-37, Col. 5 Ln. 40-46, See Figs. 1B, 4A-4B) and extracting the one or more attributes comprises identifying the one or more attributes from the image based on a visual question and answering (VQA) paradigm (Examiner interprets VQA paradigm is a known implementation of extracting attributes from images via inference and interaction. The claimed VQA-based extraction is associated to attribute inference system i.e. Yoon supplies the claimed product identification, catalog, attribute-analysis, comparison, recommendation, scoring, and ranking framework; i.e. Yoon discloses the product-identification environment, including requests directed to catalog products and user interfaces containing visually presented product items. Further the newly amended claim is disclosed by Biten for processing an input image to extract visual and textual information using a VQA architecture, where Biten’s image-based VQA front end would be used to obtain product attributes from product images for use by Yoon’s catalog-matching and recommendation system) (“with respect to a dress item, an item description may include, “this is a mini flower printed women's dress.” The attribute determination engine 110 may extract portions of the item description text. The engine 110 may then determine that the bigram “flower printed” is to be assigned a higher probability than “printed women's.” For example, the engine 110 may determine that the token “printed” appear subsequent to the token “flower” more frequently in the item descriptions 114 than “women's” appears subsequent to “printed.” The token, “mini,” may additionally be determined to have a higher probability than “mini flower.” For example, there may be a limited”) (Col. 6 ln. 58 - Col. 7 Ln. 9) wherein the VQA paradigm comprises generating (i) a question embedding, (ii) a plurality of text embeddings derived from optical character recognition (OCR) on the image, and (iii) a plurality of image embeddings generated from image patches and processing the embeddings using a neural network (Examiner notes that the underlined new amended limitation is disclosed by another prior art.)
adapting, by the processor, the one or more attributes based on historical data relating to the product into a same format as that of one or more entries in a database (Examiner interprets that extracted attributes using historical interaction data to align them with catalog attributes. This adaptation inherently requires normalizing extracted attributes into a format compatible with database entries to enable comparison) (“the attribute analysis system 100 may store the received user interactions 102 in a user interaction history database 124. The database 124 may be used by the attribute analysis system 100 to determine trend information associated with item recommendations 122. For example, the attribute analysis system 100 may learn that for a certain category of items, such as shoes, the user historically prefers certain attributes of shoes”) (Col. 8 Ln. 45-52, Col. 7 Ln. 22-36, Col. 4 Ln. 51-59), the database comprising one or more entries for each of a plurality of products (“The electronic catalog described herein may enable third-party entities to include their items in the electronic catalog. The electronic catalog may also be specific to one entity or seller. Users may access the electronic catalog via a web page or an application, and search for available items. As will be described, the web page or application may present item recommendations. The item recommendations may be determined based on monitoring prior items viewed by the users. For example, the system may determine attributes of items which are expected to be of interest to a user. In this example, the system may then identify items which include these determined attributes”) (Col. 2 Ln. 41-53, Col. 5 Ln. 40-46);
identifying a match with the product from the plurality of products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the plurality of products (Examiner interprets that a database containing multiple entries for a large plurality of products, satisfying the claimed database limitation) (“Using the mapped candidate keywords, the attribute determination engine 110 may then map similar keywords to a same common attribute. For example, it may be appreciated that different item descriptions may refer to a same attribute while using different language. With respect to a dress item, the candidates “flower print,” “floral,” “flowery pattern,” and so on, may be mapped to a same common attribute. As an example of mapping similar keywords, collaborative filtering may be used. For example, the attribute determination engine 110 may measure a frequency with which any two items mapped to different candidate keywords have been interacted with by a same set of users. Interacted with, in this specification, may include the users viewing, checking out, and so on, items via the electronic store described herein” and “For example, U.S. Pat. No. 7,752,077 discloses a process for generating an item comparison in which multiple items are compared in terms of their shared attributes, and in which the attributes are ordered or prioritized for display based on their predicted levels of importance. The attribute relevance scores disclosed herein can be used in the context of the U.S. Pat. No. 7,752,077 patent to rank or prioritize the item attributes for presentation in the item comparison”) (Col. 7 Ln. 22-36 and Col. 4 Ln. 51-59);
identifying the match with the product from the selected one or more products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the selected one or more products (Examiner interprets that the prior art compares extracted and normalized attributes against catalog entries using collaborative filtering and attribute mapping. This comparison determines which products match the extracted attributes, satisfying the matching limitation) (“attribute analysis system 100 includes an item attribute determination engine 110 which maps one or more attributes to each item included in the electronic catalog 112. As described above, the electronic catalog 112 may include item descriptions 114 for each item. These item descriptions 114 may be generated by third-party entities offering the items, and may be presented via the user interface 152 described herein. Thus, a user of the user interface 152 may view a description of an item prior to selecting the item for checkout…”) (Col. 6 Ln. 8-18, Col. 7-8);
calculating a score for each of the selected one or more products based on the comparison (“user interface 400 may include recommended items which are associated with a threshold number of the top attributes according to respective attribute relevance scores. For example, the user interface 400 may ensure that the recommended items are associated with at least 2, or 3, or 5, of the top attributes. In some embodiments, the user of the user interface 400 may specify the threshold number … The user interface 400 further includes information identifying reasons 408 for which recommended item 402 was included. The reasons 408 may identify attributes associated with the greatest attribute relevance scores. Advantageously, since the recommended items 402-406 are recommended based on specific attributes, the user interface 400 may succinctly explain reasons for the recommendations.”) (Col. 16 Ln. 30-44);
identifying the match with the product when the score for the match is a highest score of the calculated scores and greater than a threshold value (Examiner interprets that product having the highest attribute relevance scores and exceeding threshold conditions. This directly corresponds to identifying a match when a score is both highest and greater than a threshold value) (“The initial ranking may then be modified based on attribute relevance scores as described herein. For example, the user may provide a search query. In this example, the search query may relate to items or services associated with the electronic catalog. The system may obtain, or determine, items responsive to the search query. These items may then be ranked, for example according to popularity, the user-preference information, and so on. Attributes associated with the ranked items may then be identified. For example, the attributes associated with a top threshold number of the items may be identified. Attribute relevance scores may then be determined for the attributes. The items may then be-ranked based on the attribute relevance scores. As an example, the system may increase rankings of items with greater numbers of attributes having the greatest attribute relevance scores”) (Col. 13 Ln. 52-67, Col. 10-11, Col. 16 Ln. 30-44).
Yoon specifically doesn’t discloses, selecting one or more products from the plurality of products in the database based on a rule-based comparison of at least one of the one or more attributes with each corresponding entry in the database, verifying whether each of the selected one or more products excluding the identified match having a score greater than the threshold value is also a match with the product, and retraining the processor based on the verification, however Krasadakis discloses, selecting one or more products from the plurality of products in the database based on a rule-based comparison of at least one of the one or more attributes with each corresponding entry in the database (Examiner interprets that uses rule-based comparisons (e.g., thresholds, parameters, elasticity rules) to select candidate products, satisfying the rule-based selection i.e. Krasadakis discloses evaluating seller-product offerings using predefined buyer and seller parameters and elasticity rules. The evaluated parameters include price ranges, percentages of desired product specifications, availability periods, social-media indicators, and other defined conditions. Candidate buyer-retailer pairings satisfying the applicable threshold conditions are identified and used to create a list of qualifying retailer product offerings.) (“taking the predefined buyer and seller parameters and elasticity and identifying buyer-retailer pairings that facilitate the best product transactions. For example, a multitude of seller AI negotiation agents offering the desired products being searched for by an AI negotiation agent within elasticity thresholds of the various of buyer parameters (e.g., within 10% of price, having 90% of the desired product specifications, with an availability within hours or a day of the buyer's request, being mentioned a certain number of times in social media, etc.) cause some of the disclosed examples to create a list of retailer product offerings to present to the buyer, or may automatically purchase the products from one of the sellers in view of the buyer's preset authorization …”) (0025-0027);
verifying whether each of the selected one or more products excluding the identified match having a score greater than the threshold value is also a match with the product (Examiner interprets that the prior art evaluates multiple candidate offerings that meet threshold criteria before final selection, which corresponds to verifying whether additional products above a threshold are also valid matches i.e. Krasadakis further discloses evaluating multiple candidate seller-product offerings against predefined threshold criteria to determine whether each offering qualifies for inclusion in the candidate list. Thus, in addition to determining a preferred transaction, Krasadakis evaluates whether other candidate offerings satisfy the applicable buyer parameters and remain qualifying product matches) (“taking the predefined buyer and seller parameters and elasticity and identifying buyer-retailer pairings that facilitate the best product transactions. For example, a multitude of seller AI negotiation agents offering the desired products being searched for by an AI negotiation agent within elasticity thresholds of the various of buyer parameters (e.g., within 10% of price, having 90% of the desired product specifications, with an availability within hours or a day of the buyer's request, being mentioned a certain number of times in social media, etc.) cause some of the disclosed examples to create a list of retailer product offerings to present to the buyer, or may automatically purchase the products from one of the sellers in view of the buyer's preset authorization …”) (0025-0028)
and retraining the processor based on the verification (Examiner interprets that the cited art adaptively updates parameters based on outcomes and performance feedback. Such adaptive updating constitutes retraining of the processor to improve future selections i.e. the existing position that elasticity ranges and system parameters are adaptively set or updated based on market conditions, availability, reviews, trends, performance, and completion rates, thereby adapting future product selections i.e. Krasadakis additionally teaches adaptively setting or updating the elasticity parameters used by its AI buyer and seller agents based on market conditions, product availability, trends, reviews, overall performance, completion rates, and other feedback) (“Elasticity may be set by the buyer and seller themselves, by the current market conditions, by product availability, by the seasonality or trends of products, by social media or online commentary, by product reviews, or a combination thereof or by the overall performance and completion rate in reference to the strategic goals and objectives. For instance, buyers and sellers may specify particular elasticity ranges for their respective AI buyer and seller agents”) (0021, 0019-0020).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention for extracting, by a processor, one or more attributes relating to a product from data indicated in a request for identifying the product, each of the one or more attributes describing the product, wherein the data comprises an image of the product and extracting the one or more attributes comprises identifying the one or more attributes from the image based on a visual question and answering (VQA) paradigm; adapting, by the processor, the one or more attributes based on historical data relating to the product into a same format as that of one or more entries in a database, the database comprising one or more entries for each of a plurality of products; identifying a match with the product from the plurality of products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the plurality of products, identifying the match with the product from the selected one or more products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the selected one or more products; calculating a score for each of the selected one or more products based on the comparison; identifying the match with the product when the score for the match is a highest score of the calculated scores and greater than a threshold value, as disclosed by Yoon, selecting one or more products from the plurality of products in the database based on a rule-based comparison of at least one of the one or more attributes with each corresponding entry in the database, verifying whether each of the selected one or more products excluding the identified match having a score greater than the threshold value is also a match with the product, and retraining the processor based on the verification, as taught by Krasadakis for the purpose to use the results of the candidate verification process as feedback for updating the matching parameters or model executed by the processor thus causing the candidates previously verified as acceptable or unacceptable matches to influence subsequent candidate selection and matching, thereby improving later product-identification results.
Yoon specifically doesn’t discloses, wherein the VQA paradigm comprises generating (i) a question embedding, (ii) a plurality of text embeddings derived from optical character recognition (OCR) on the image, and (iii) a plurality of image embeddings generated from image patches and processing the embeddings using a neural network, however Biten discloses, wherein the VQA paradigm comprises generating (i) a question embedding (Examiner interprets Biten discloses receiving a question as one of three input modalities. The question is tokenized into question tokens Q={W1,…,Wm}, and the question tokens are embedded to produce encoded question features (Eq). The encoded question features are subsequently provided to the multimodal transformer architecture i.e. it then uses those features as one of the three inputs to the multimodal transformer) (“LaTr So far, we explained the building blocks of our method, now we describe how we put it all together, as depicted in Fig. 2 (b). After pre-training the language modality of the model with layout information, we input all three modalities, namely; image, OCR information and question to the transformer encoder. Let V = {V₀, ..., VN } be a set of visual patch features such that V₀ is the [class] embed-ding, Q = {W₁, ..., Wm} be the question tokenized into Wi and O = {O₁, O₂, ..., On} be the OCR tokens. We embed the OCR tokens and questions using Eq. (1) to obtain encoded OCR tokens E and encoded question features E q. For the 2-D spatial embedding of each Wi, we use fixed values (x₀ = y₀ = 0; x₁ = y₁ = 1000). Finally, we concatenate all the inputs [V; E ; E q] to feed to the multimodal trans-former encoder-decoder architecture. Cross entropy loss is used to fine-tune our model”) (Page 4, “LaTr.”),
(ii) a plurality of text embeddings derived from optical character recognition (OCR) on the image (Examiner interprets Biten discloses obtaining OCR tokens from an image and embedding the OCR tokens to produce encoded OCR-token representations E. Figure 2 shows an “OCR System” followed by “Word Embedding,” and Figure 3 identifies the resulting semantic representations as “Text Embeddings.” Biten further explains that an OCR token and its associated spatial information are combined into an encoded representation before being supplied to the transformer encoder.) (“Formally, as seen in Fig. 3, given an OCR token Oi, the associated word bounding box may be defined by (xi0; yi0; xi1; yi1; hi; wi), where (xi0; yi0) corresponds to the position of the upper left corner of the bounding box,(xi1; yi1) represents the position of the lower right corner, and (hi;wi) represents the height and width with respect to the reading order. To embed bounding box information, we use a lookup table commonly used for continuous encoding one-hot representations (e.g. nn.Embedding in PyTorch). Before we feed the word representation into the transformer encoder, we sum up all the representations together: … Most previous methods utilized an external pre-trained object detector [20, 70] for extracting ob-jects labels, visual object features and visual OCR features. In this work, we diverge from the literature and leverage a Vision Transformer (ViT) [12]. The ViT is an image clas-sification network which is pre-trained and fine-tuned on ImageNet [10]. We utilize ViT in our architecture only in the fine-tuning stage, and we freeze all the layers except the last fully connected projection layer we add. Formally, an image I having the dimension of H _W _ C is reshaped nto 2D patches of size N _ (p2 _ C), where (H;W) is the height and width, C is the number of channels, (P; P) is the resolution of each image patch, and N = HW=P2 is the final number of patches.”) (Page 3-4, Figs. 2-3),
and (iii) a plurality of image embeddings generated from image patches (Examiner interprets that Biten discloses using a Vision Transformer to generate visual features. The input image is reshaped into a plurality of two-dimensional image patches. The flattened patches are mapped into a multidimensional embedding space using a linear projection layer, and positional embeddings are added to the patch embeddings. The resulting visual output is represented as a sequence of visual patch features (V={VO, …, VN) (“Visual Features Most previous methods utilized an external pre-trained object detector [20, 70] for extracting objects labels, visual object features and visual OCR features. In this work, we diverge from the literature and leverage a Vision Transformer (ViT) [12]. The ViT is an image classification network which is pre-trained and fine-tuned on ImageNet [10]. We utilize ViT in our architecture only in the fine-tuning stage, and we freeze all the layers except the last fully connected projection layer we add. Formally, an image I having the dimension of H _W _ C is reshaped nto 2D patches of size N _ (p2 _ C), where (H;W) is the height and width, C is the number of channels, (P; P) is the resolution of each image patch, and N = HW=P2 is the final number of patches.”) (Page 4, Figs. 2-3)
and processing the embeddings using a neural network (Examiner interprets that Biten discloses concatenating the visual patch features V, encoded OCR-token features E, and encoded question features Eq ([V; E; Eq]). The concatenated representations are then supplied to a multimodal transformer encoder-decoder architecture. The transformer encoder-decoder constitutes the claimed neural-network processing of the embeddings) (“Visual Features Most previous methods utilized an external pre-trained object detector [20, 70] for extracting objects labels, visual object features and visual OCR features. In this work, we diverge from the literature and leverage a Vision Transformer (ViT) [12]. The ViT is an image classification network which is pre-trained and fine-tuned on ImageNet [10]. We utilize ViT in our architecture only in the fine-tuning stage, and we freeze all the layers except the last fully connected projection layer we add. Formally, an image I having the dimension of H _W _ C is reshaped nto 2D patches of size N _ (p2 _ C), where (H;W) is the height and width, C is the number of channels, (P; P) is the resolution of each image patch, and N = HW=P2 is the final number of patches … method, now we describe how we put it all together, as depicted in Fig. 2 (b). After pre-training the language modality of the model with layout information, we input all three modalities, namely; image, OCR information and question to the transformer encoder. Let V={VO, …, VN) be a set of visual patch features such that V0 is the [class] embedding, Q = {W1, …, WM) be the question tokenized into Wi and O = {O1;O2; … ;On} be the OCR tokens. We embed the OCR tokens and questions using Eq. (1) to obtain encoded OCR tokens E and encoded question features Eq. For the 2-D spatial embedding of each Wi, we use fixed values (x0 = y0 = 0; x1 = y1 = 1000). Finally, we concatenate all the inputs [V; E; Eq] to feed to the multimodal transformer encoder-decoder architecture. Cross entropy loss is used to fine-tune our model.”) (Page 4, Figs. 2-3).
It would have been obvious to a person of ordinary skill in the art before the effective filling date of the applicant’s invention for extracting, by a processor, one or more attributes relating to a product from data indicated in a request for identifying the product, each of the one or more attributes describing the product, wherein the data comprises an image of the product and extracting the one or more attributes comprises identifying the one or more attributes from the image based on a visual question and answering (VQA) paradigm; adapting, by the processor, the one or more attributes based on historical data relating to the product into a same format as that of one or more entries in a database, the database comprising one or more entries for each of a plurality of products; identifying a match with the product from the plurality of products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the plurality of products, identifying the match with the product from the selected one or more products, the identification based on comparing, by the processor, the one or more attributes with each corresponding entry of the selected one or more products; calculating a score for each of the selected one or more products based on the comparison; identifying the match with the product when the score for the match is a highest score of the calculated scores and greater than a threshold value, as disclosed by Yoon, wherein the VQA paradigm comprises generating (i) a question embedding, (ii) a plurality of text embeddings derived from optical character recognition (OCR) on the image, and (iii) a plurality of image embeddings generated from image patches and processing the embeddings using a neural network, as taught by Biten for the purpose to apply multimodal VQA technique to the product images used in Yoon’s catalog environment to obtain both visual and textual product attributes for subsequent catalog comparison and matching, thus by applying Biten’s architecture to extract image based attributes to normalize, compare, score, and match those attributes against catalog entries.
As per claims 2 and 6, Yoon discloses, wherein selecting the one or more products from the plurality of products in the database based on the rule-based comparison further comprises: identifying one or more key attributes from the one or more attributes (Examiner interprets that cited art identifies and prioritizes attributes using relevance scoring. Selecting “key attributes” is inherent in ranking attributes by importance, and selecting products based on those key attributes) (“attribute analysis system 100 includes an item attribute determination engine 110 which maps one or more attributes to each item included in the electronic catalog 112. As described above, the electronic catalog 112 may include item descriptions 114 for each item. These item descriptions 114 may be generated by third-party entities offering the items, and may be presented via the user interface 152 described herein. Thus, a user of the user interface 152 may view a description of an item prior to selecting the item for checkout…”) (Col. 6 Ln. 8-18, Col. 7-8);
and selecting the one or more products from the plurality of products in the database based on comparing the one or more key attributes with each corresponding entry in the database (“Using the mapped candidate keywords, the attribute determination engine 110 may then map similar keywords to a same common attribute. For example, it may be appreciated that different item descriptions may refer to a same attribute while using different language. With respect to a dress item, the candidates “flower print,” “floral,” “flowery pattern,” and so on, may be mapped to a same common attribute. As an example of mapping similar keywords, collaborative filtering may be used. For example, the attribute determination engine 110 may measure a frequency with which any two items mapped to different candidate keywords have been interacted with by a same set of users. Interacted with, in this specification, may include the users viewing, checking out, and so on, items via the electronic store described herein” and “For example, U.S. Pat. No. 7,752,077 discloses a process for generating an item comparison in which multiple items are compared in terms of their shared attributes, and in which the attributes are ordered or prioritized for display based on their predicted levels of importance. The attribute relevance scores disclosed herein can be used in the context of the U.S. Pat. No. 7,752,077 patent to rank or prioritize the item attributes for presentation in the item comparison”) (Col. 7 Ln. 22-36 and Col. 4 Ln. 51-59).
As per claims 3 and 7, Yoon discloses, further comprising adapting, by the processor, the one or more attributes into at least one of a unit of measurement, quantity, language or item name that is the same as that of one or more entries in the database (Examiner interprets the cited art disclose to normalize semantically equivalent attributes (e.g., “flower print,” “floral”) into common representations. Such normalization necessarily includes adapting language, naming conventions, and units to match database formats) (“Using the mapped candidate keywords, the attribute determination engine 110 may then map similar keywords to a same common attribute. For example, it may be appreciated that different item descriptions may refer to a same attribute while using different language. With respect to a dress item, the candidates “flower print,” “floral,” “flowery pattern,” and so on, may be mapped to a same common attribute. As an example of mapping similar keywords, collaborative filtering may be used. For example, the attribute determination engine 110 may measure a frequency with which any two items mapped to different candidate keywords have been interacted with by a same set of users. Interacted with, in this specification, may include the users viewing, checking out, and so on, items via the electronic store described herein”) (Col. 7 Ln. 22-36 and Col. 4 Ln. 51-59).
As per claims 4 and 8, Yoon discloses, wherein the database further indicates an identifier for each of the plurality of products (“electronic catalog described herein may enable third-party entities to include their items in the electronic catalog. The electronic catalog may also be specific to one entity or seller. Users may access the electronic catalog via a web page or an application, and search for available items. As will be described, the web page or application may present item recommendations. The item recommendations may be determined based on monitoring prior items viewed by the users. For example, the system may determine attributes of items which are expected to be of interest to a user. In this example, the system may then identify items which include these determined attributes. One or more of these identified items may then be recommended to the user”) (Col. 2 Ln. 41-54, Fig. 5, Col. 6), the method further comprising mapping the product with the identifier of the identified match (Examiner interprets electronic catalog inherently associates products with identifiers (e.g., catalog entries). Selecting and recommending a product necessarily maps the product to its identifier) (“FIG. 1B is an example user interface 160 illustrating recommended items 162 according to the techniques described herein. The example user interface 160 may represent an embodiment of the user interface 152 in which the user is searching the electronic catalog 112 for hats. For example, the user may optionally provide a search query or request to find hats. As another example, the user may select which are commonly viewed, or selected for checkout, by other users”) (Col. 11 Ln. 29-37, See Figs. 1B, 4A-4B).
Response to Arguments
With regards to § 101 rejections:
The arguments filed on May 12th, 2026, with respect to the rejection(s) of claims 1-8 under 35 U.S.C 101 have been fully considered but are unpersuasive.
Applicant states that the amended claims recite a specific VQA architecture and a two-stage framework that reduces computational complexity. Applicant's arguments have been fully considered but they are not persuasive. The claims remain focused on collecting product information, extracting and normalizing attributes, comparing the attributes with database entries, filtering and scoring candidates, selecting a matching product, verifying additional candidates, and retraining based on the results. Considered as a whole, these limitations recite the abstract idea of analyzing and evaluating information to make a product identification decision (Under Step 2A, Prong One). The amended VQA limitations do not establish the asserted improvement to computer functionality. Although the claims require question, OCR text, and image patch embeddings processed by a neural network, they do not recite a particular neural network configuration, fusion technique, training procedure, or technical mechanism that improves the operation of the computer itself (Under Step 2A, Prong Two). Moreover, the alleged reduction in computational complexity is not commensurate with the claim scope. The neural network is used during attribute extraction, whereas the later rule based filtering and candidate scoring steps do not require filtering candidates before neural network processing. The claims therefore do not require the sequence alleged to produce the stated efficiency improvement. Further, the claims differ from Enfish because they do not recite an improved computer data structure or a change in how the computer operates. Instead, the computer components implement the product matching analysis at a functional level. A technological improvement argument must be reflected in the claim itself and evaluated in view of the specification; merely limiting an abstract process to AI or neural network technology is insufficient. The processor, database, OCR processing, embeddings, neural network, comparison rules, scoring, and retraining are recited as tools for carrying out the underlying information analysis process. The ordered combination does not require an unconventional technical arrangement beyond implementing the product identification workflow on computer technology. Accordingly, the amended claims remain directed to a judicial exception without additional elements that integrate the exception into a practical application or amount to significantly more, and therefore, the §101 rejection is maintained. See MPEP §§ 2106.05(a)–(c), (e)– (h).
With regards to § 103 rejections:
Applicant's arguments, see pages 8-9, filed May 12th, 2026, with respect to the rejection(s) of claims 1-8 under 35 U.S.C 102/103 have been fully considered but are unpersuasive/moots on new ground of rejection.
Applicant states that Yoon merely extracts portions of textual item descriptions using bigrams and does not extract attributes from an image or disclose the specific VQA mechanics now recited in claim 1. Applicant's arguments have been fully considered but they are not persuasive, however Applicant’s statement is acknowledged to the extent that Yoon, standing alone, does not expressly disclose the newly added question embedding, OCR text embedding, and image patch embedding architecture. The present rejection does not rely on Yoon alone for those limitations. Yoon is relied upon for the underlying product identification and catalog matching process, including product attributes, catalog entries, comparisons, relevance scoring, ranking, and identification of responsive products. Biten is separately relied upon for extracting information from an image using the claimed multimodal VQA architecture. The proposed modification does not equate Yoon’s bigram analysis with the claimed VQA architecture. Rather, Biten’s VQA architecture is incorporated as the image-based attribute extraction mechanism supplying attributes to Yoon’s product matching process. Applicant’s original statement accurately identified the deficiency in the former reliance on Yoon, but Biten directly addresses that amended subject matter.
Further, Applicant states that the prior rejection failed to identify where Yoon or Krasadakis teaches the specific combination of question, OCR text, and image patch embeddings processed by a neural network. Applicant's arguments have been fully considered but they are not persuasive, however the current amended claims does not overcome the present rejection because the newly applied Biten reference expressly discloses that combination. Biten specifically discloses, tokenizes and embeds the question to produce encoded question features; embeds OCR tokens obtained from the image to produce encoded OCR-text features; divides the image into patches and projects the patches into visual feature representations; and concatenates the visual, OCR-text, and question representations and supplies them to a multimodal transformer encoder-decoder, See Biten pages 3-4, Figs. 2-3.
Further, Applicant states that Krasadakis merely adjusts price or negotiation parameters in response to market conditions and does not verify “confusing cases” or retrain a model based on the verification. Applicant's arguments have been fully considered but they are not persuasive to the extent it relies on limitations not present in claim 1. Claim 1 does not recite: “confusing cases”; Claim 1 more broadly requires verifying whether the remaining selected products above the threshold are also matches and “retraining the processor based on the verification.” Krasadakis discloses, evaluating multiple product offerings against predefined threshold criteria and identifying those offerings that satisfy the buyer’s requirements. Krasadakis further discloses adaptively setting the parameters used by its AI agents based on performance, completion rates, reviews, market conditions, and other feedback. In the proposed combination, the qualification results obtained from verifying Yoon’s remaining above threshold candidates would constitute performance feedback used to update the programmed matching parameters or model executed by the processor. Using the results of candidate verification as the feedback input for Krasadakis’s adaptive updating would have been a predictable use of the disclosed adaptive technique to allow verified matching outcomes to affect subsequent matching decisions and refine future product selection.
Applicant further states that the claimed invention uses a feedback loop to improve model accuracy for difficult matches, whereas Krasadakis merely tunes negotiation parameters. Applicant's arguments have been fully considered but they are not persuasive, as claim 1 does not specify that the verified candidates are “difficult matches,” that a model-accuracy metric is calculated, or that any particular neural-network weights are modified. Nor does claim 1 identify the training procedure, training objective, model architecture being retrained, or manner in which the verification changes the model. Under the broadest reasonable interpretation, “retraining the processor” encompasses updating or retraining the programmed matching logic or AI model executed by the processor based on the verification results. If Applicant intends to require retraining of a particular neural-network model using particular verified examples, those restrictions are not presently recited in the claim. Krasadakis’s adaptive updating based on system performance, when applied to Yoon’s verified candidate outcomes, therefore meets or at least renders obvious the broadly recited feedback based retraining limitation.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. US. Pat. 10949907 (“Jain”).
Jain discloses, a method for generating a list of products each matching a reference product are disclosed. A user query is first received, and multi-modal attribute data for the reference product are determined, with each data mode being a type of product characterization having a modality selected from a text data class, categorical data, a pre-compared engineered feature, audio, image, and video. Next, a first list of candidate products is determined based on a product match signature, and a second list of candidate products is generated from the first, wherein for at least one given candidate product, a deep learning multi-modal matching model is selected to determine whether a match is found. Lastly, the second list is filtered to remove outliers and to generate the list of matching products.
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to 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 at (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