DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
This Office action is in response to correspondence received May 15, 2026.
Claims 1 and 11 are amended. 1-20 are pending and have been examined.
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 an abstract idea without significantly more. The claim(s) recite(s):
Claim 1:
processing consumer reviews, access the consumer reviews associated with a service provider; select at least one of the consumer reviews which is relevant to at least one predetermined category; obtain an annotation associated with the selected consumer review with a third party; generate a tag content associated with the selected consumer review by summarizing the selected consumer review; and generate a tag associated with the selected consumer review based on the tag content and the annotation associated with the selected consumer review, classify the selected consumer review based on at least one property of the selected consumer review, and distribute a task for the annotation associated with the selected consumer review with the third party based on the classification of the selected consumer review.
Claim 11:
A method for processing consumer reviews, the method comprising: accessing the consumer reviews associated with a service provider; selecting at least one of the consumer reviews which is relevant to at least one predetermined category; classifying the selected consumer review based on at least one property of the selected consumer review; distributing a task for an annotation associated with the selected consumer review with a third party based on the classification of the selected consumer review; obtaining the annotation associated with the selected consumer review with the third party; generating a tag content associated with the selected consumer review by summarising the selected consumer review; and generating a tag associated with the selected consumer review based on the tag content and the annotation associated with the selected consumer review.
Claims 1 and 11 recite an abstract idea that is a commercial interaction, which is a certain method of organizing human activity. The steps recite a way for consumers to rate and review a business and then the review is further annotated and tagged, in other words, adding more information to it. The process of generating and further detailing reviews is marketing or sales activities or behaviors because it is used to promote a business, and marketing or sales activities or behaviors are one of the commercial interactions described in MPEP 2106.04(a). Moreover, consumer reviews and these specific limitations above are specific human activity that is being organized, which is managing personal behavior or relationships or interactions between people (including social activities, teaching, and following rules or instructions). See Id. Therefore for these reasons claims 1 and 11 recite a certain method of organizing human activity.
This judicial exception is not integrated into a practical application. The claims recite generic computing components that alone and in combination amount to performing the steps on a computer or other machinery in its ordinary capacity. This includes using a server, see TLI Communications, MPEP 2106.05(f)(2). The additional elements are:
Claim 1:
A server for
the server comprising: a memory for storing instructions; and a processor for executing the stored instructions and configured to/further configured to:
using a natural language generation engine
obtain from a computing device associated; distribute to the computing device associated
Claim 11:
to a computing device associated/from the computing device associated
using a natural language generation engine
In combination these amount to no more than a computer receiving/displaying (distributing) information and exchanging information with a server and using a natural language generation engine in an apply it manner. This to perform the steps of the abstract idea above. This amounts to performing a commonplace business practice on a computer, which is an apply it limitation and not a practical application. See A commonplace business method or mathematical algorithm being applied on a general purpose computer, Alice Corp. Pty. Ltd. V. CLS Bank Int’l, 573 U.S. 208, 223, 110 USPQ2d 1976, 1983 (2014); Gottschalk v. Benson, 409 U.S. 63, 64, 175 USPQ 673, 674 (1972); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015), See MPEP 2106.05(f)(2). Taking the claims as a whole, this is no more than an abstract idea (identified above) with applied generic computer elements and applied algorithm to the abstract idea. Therefore the additional elements alone and in combination do not integrate the abstract idea into a practical application.
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the reasoning above is carried over here: for the same reason that the combination of elements is not a practical application, it is not significantly more than the abstract idea.
Per the dependent claims:
The dependent claims, 2-10 and 12-20, further describe the abstract idea of claims 1 and 11 and therefore do not overcome the 101 rejection. Claims 6 and 16 recite using a natural language processing model but this is an apply it limitation that is not a practical application or significantly more, either alone or in combination.
Therefore, claims 1-20 are rejected under 35 USC 101.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-5, 8-15, and 18-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Parveen et al., US PGPUB 20170068999 A1 ("Parveen") in view of Deluca et al., US PGPUB 20190180353 A1 ("Deluca"), further in view of Nidell et al., US PGPUB 20210166282 A1 (“Nidell”).
Per claims 1 and 11, which are similar in scope, Parveen teaches A server for processing consumer reviews, the server comprising: a memory for storing instructions in par 27: “In one embodiment, a system can comprise: one or more processing modules; and one or more non-transitory storage modules storing computing instructions configured to run on the one or more processing modules and perform the acts of:”
Parveen then teaches and a processor for executing the stored instructions and configured to: access the consumer reviews associated with a service provider in par 29: “and one or more non-transitory storage modules storing computing instructions configured to run on the one or more processing modules and perform the acts of analyzing one or more reviews in a set of reviews to determine a set of features”
See also par 040 for service providers: “The Internet gives users the ability to research any product or service that the user wishes to purchase to a much greater extent than possible before the Internet. For example, users can go to an eCommerce website, search for a product or service, and analyze the opinions of many different people regarding the product or service. This ability can be particularly useful for users who are comparing two competing products or services.”
Parveen then teaches select at least one of the consumer reviews which is relevant to at least one predetermined category in par 53: “Thereafter, for every feature in the feature set, the user-generated content (“UGC”) is analyzed to find mentions of each feature (block 304). Thereafter, each of the mentions can be analyzed to determine polarity of each mention (block 306). “ The user generated content teaches consumer reviews. The features are the categories.
See also par 56: :For a feature f.sub.i in the feature set, the number of occurrences in review R.sub.j is n.sub.ij.”
See also par 060. See also Par 050, initial feature set, which as it is determining initial features, teaches predetermined categories (drawn from the specification of the product). See also par 052.
Parveen then teaches obtain an annotation associated with the selected consumer review from a computing device associated with a third party in par 95: “For example, another dimension is added to reviews. Annotating the reviews with relevant features of the product makes the reviews more readable for users. Thus, information is more accessible to users as the user can read only the reviews that discuss the features of interest to him/her.”
See also par 99: “As disclosed above, there can be methods and systems to calculate a feature score for every feature of a product. The feature score can be used to annotate reviews for which the feature score is not equal to zero. If a feature has either a positive feature score or a negative feature score, the implication is that the review mentions that particular feature. It is also possible for a review to have a neutral feature score even if the feature is mentioned in the review.” Feature score is the annotation obtained.
See also par 103: “With reference to FIG. 9, a flowchart illustrating a method 900 for annotating reviews is presented. The review annotation can be performed on a computer system such as computer system 100 (FIG. 1). Method 900 is merely exemplary and is not limited to the embodiments presented herein. Method 900 can be employed in many different embodiments or examples not specifically depicted or described herein. In some embodiments, the procedures, the processes and/or the activities of method 900 can be performed in the order presented. In other embodiments, the procedures, the processes, and/or the activities of method 900 can be performed in any other suitable order. In still other embodiments, one or more of the procedures, the processes, and/or the activities of method 900 can be combined or skipped.” Computer system 100 teaches computing device associated with a third party.
Then, Parveen teaches generate a tag content associated with the selected consumer review by summarizing the selected consumer review in par 114: “Column 1020 contains a list of “tags” that represent features that received either positive or negative feature scores in each review. In some embodiments, the tags can be color-coded. In such a manner, one color can represent positive feature scores and another color can represent negative feature scores. Instead of color coding, some embodiments can use other methods of indicating the difference between positive, negative, and neutral feature scores, such as shading, hatch marks, underlining, typeface, font colors, and the like.” See for generated par 0107: “And the features with a feature score less than a given threshold score (e.g., zero), can get marked as “Bad.” Then the reviews can be displayed to the user with a tag indicating the feature and whether the feature score was “Good” or “Bad” (block 910). Displaying a review to the User can involve transmitting data to a user's computer that cause the user's computer to display reviews.”
Then Parveen teaches and generate a tag associated with the selected consumer review based on the tag content and the annotation associated with the selected consumer review in par 116: “In some embodiments, a user is able to click on a tag to view only those reviews that contain features scores for that feature. Moving ahead to FIG. 11, a screen shot 1100 is presented. FIG. 11 is merely exemplary and embodiments of the screen representation are not limited to the embodiments presented herein. The screen representation can be employed in many different embodiments or examples not specifically depicted or described herein.”
See also par 118: “Column 1110 displays the title of the review. In some embodiments, only the tag that is the subject of the search is displayed to the user in column 1120. In some embodiments, all of the pertinent tags can be displayed to the user, to allow the user to view other feature scores. In some embodiments, instead of a tag being displayed in column 1120, because all of the reviews are relevant to a selected feature, an indication of whether the review is positive or negative can be contained in column 1120.”
Then Parveen teaches wherein the processor is further configured to classify the selected consumer review based on at least one property of the selected consumer review in par 120: “Pie chart 1150 can contain various segments, such as positive segment 1160, neutral segment 1162, and negative segment 1164. Each of these segments is configured to illustrate the percentage of reviews that contain a certain feature score. In FIG. 11, the vast majority of reviews that contain a feature score for “body” are positive, so positive segment 1160 is much larger than the other two segments. In some embodiments, there can be color-coding of the graphical indicia. For example, positive segment 1160 can be a first color, neutral segment 1162 can be a second color, and negative segment 1164 can be a third color. The color-coding can extend to the tags in column 1120. For example, a review that contained a positive feature score for “body” can be the same first color as positive segment 1160. In a similar manner, a review that contained a neutral feature score for “body” can be the same second color as neutral segment 1162 and a review that contained a negative feature score for “body” can be the same third color as negative segment 1164. While a pie chart is illustrated in FIG. 11, other embodiments can use other types of graphical indicia, such as bar graphs, line graphs, histograms, and the like.” Under a broadest reasonable interpretation in light of the specification, the pie graph or other graph teaching here teaches classifying the consumer review based on at least one property of the selected consumer review because it is taking the review and giving it the classification of overall good or bad.
Parveen does not teach and distribute a task for the annotation associated with the selected consumer review to the computing device associated with the third party based on the classification of the selected consumer review.
Deluca teaches normalizing product reviews and the presentation of product information viewed within a vendor's virtual storefront. See abstract.
Deluca teaches and distribute a task for the annotation associated with the selected consumer review to the computing device associated with the third party based on the classification of the selected consumer review in par 56: “In some instances however, a user who is seeking to normalize product reviews to a target user may manually input characteristics and information into the data collection module 111 about the target user of interest. For example, a user seeking to normalize product reviews to one of their siblings may input information about that sibling including the sibling's age, sizing information, location or environment within which their sibling resides, activity level, past purchases or trends the sibling follows. Moreover, in some embodiments, the data collection module 111 may also automatically or manually be directed to search one or more network data sources 135 for information about a target user.” The system taught by Deluca has the user annotate which teaches distribute a task.
Then see par 69: “In some embodiments of the algorithm 500, the normalized product reviews 423 may be annotated by the annotation module 115 in step 522. One or more of the normalized product reviews 423 may include one or more annotations 425 highlighting or depicting relevant portions of the product reviews most relevant to the target user. The annotation module 115 and/or prioritization module 117 may also generate a summary 421 for each product viewed by a user experiencing the vendor's virtual storefront, wherein the summary 421 may provide reasons, rationales and keywords or phrases associated with the target user. Each of the reasons, rationales and keywords may be identified and described to the user in a manner that assists the user with understanding the reasoning for one or more normalized product reviews 423 to be prioritized over some of the other product reviews which may have been rated as a lower priority or disregarded as being irrelevant to the target user for one reason or another.”
It would have been obvious to one ordinarily skilled in the art before the effective filing date of the claimed invention to modify the annotation and tagging of reviews teaching of Parveen with the distributing a task teaching of Deluca because Deluca teaches in par 018 that the annotations can be designed for the benefit of the target user, as taught further in par 017 the user’s interests, purchasing history, etc. As this would make reviews more relevant (The teaching above is for the annotations that would be performed in this way), one would be motivated to combine Deluca with Parveen so that a potential user who is reading the review content gets more relevant information. For these reasons, one would be motivated to modify Parveen with Deluca.
Parveen does not teach generate a tag content associated with a consumer review using a natural language generation engine by summarising the selected consumer review.
Nidell teaches using AI to dynamically determine sub-category ratings from content commentary of reviews of an online reviews forum. See abstract.
Nidell teaches generate a tag content associated with a consumer review using a natural language generation engine by summarising the selected consumer review in par 048: “At step (304), the review counting variable X is initialized, which corresponds to the first review or “Review 1.” At step (306), natural language processing (NLP) is applied to the content commentary of the accessed review, e.g., Review.sub.X, to generate machine-readable sub-topic data and machine-readable sentiment data. In some embodiments, the machine-readable data comprises vectors, e.g., word, phrase, sentence, document, other vector representations, or combinations thereof. At step (308), artificial intelligence (AI) is applied to the sub-topic data and the sentiment data to dynamically identify a sub-topic category or a plurality of sub-topic categories (e.g., food quality, affordability, service, etc.) from the sub-topic data, and the quantity of sub-topic categories is assigned to the value Y.sub.Total. For example, if Review.sub.X discusses food quality and service only, then the A will dynamically identify two sub-topic categories, so that Y.sub.Total will equal 2.” Sub-topic and sentiment data teach content tag.
It would have been obvious to one ordinarily skilled in the art before the effective filing date of the claimed invention to modify the content annotation teaching of Parveen with the generation of tags by natural language teaching for content tagging teaching of Nidell because Nidell teaches in par 005 that: “The disparities in sub-topic prioritization by reviewing entities is not taken into consideration in computing the overall rating, and more importantly leads to overall ratings that are based on sub-topic prioritizations of reviewing entities that do not align with the priorities of the searching entities accessing the review forum. For example, a searching entity may have a health condition that makes ease of accessibility to a given venue, such as a restaurant, the highest priority for the searching entity. On the other hand, the overall rating may be computed from underlying overall ratings of reviewing entities who place priority on sub-topics other than accessibility (e.g., cleanliness, food quality, etc.). The average overall rating computed from those overall ratings may not reflect the topic priority or priorities of the searching entity. Further, browsing through the tens, hundreds, or even thousands of personalized comments for a specific sub-topic that is of importance to the searching entity, such as accessibility, can be extremely time-consuming and inefficient for the searching entity.” This way of evaluating reviews prevents a reviewing entity bias that would distort the review summary, making it less useful for people reading the reviews. Because Nidell teaches this advantage, one would be motivated to modify Praveen with Nidell in order to have more balanced summary information of reviews.
Per claims 2 and 12, which are similar in scope, Parveen, Deluca, and Nidell teach the limitations of claims 1 and 11, above. Parveen does not teach wherein the processor is further configured to update a rule for selecting the at least one of the consumer reviews which is relevant to the at least one predetermined category, based on the annotation obtained from the computing device associated with the third party.
Deluca teaches wherein the processor is further configured to update a rule for selecting the at least one of the consumer reviews which is relevant to the at least one predetermined category, based on the annotation obtained from the computing device associated with the third party in par 70: “In step 523 of the algorithm 500, the vendor management system 103 may display to the display device 110 of the client device 131 the product data 401 selected by the user, the normalized reviews 423 prioritized based on the target user's profile 121, preferences, habits and characteristics as well as any annotations 425 and/or summaries 421 describing the reasons and rationale for normalizing the product reviews 425 in the order presented. The algorithm may subsequently proceed to step 525 to determine if the user has completed the navigation of the vendor′ virtual storefront on behalf of the target user and/or has the user completed a transaction for purchasing one or more of the products or services viewed in step 523. If the navigation of the vendor's virtual storefront using the target user's profile 121 has not concluded and/or a final transaction (such as a purchase) has not been completed, the algorithm may proceed to step 519, wherein the user may continue to navigate the product and service offerings of the vendor's virtual storefront.”
See also par 65 “In step 505 however, the user may transmit a request to the profile module 107 instructing the profile module 107 to load a customized virtual storefront normalized to a target user's profile 121, a set of manually inputted target user characteristics or by specifying one or more third party profiles (such as social media) maintained by a network data source 135.”
It would have been obvious to one ordinarily skilled in the art before the effective filing date of the claimed invention to modify the annotation and tagging of reviews teaching of Parveen with the distributing a task teaching of Deluca because Deluca teaches in par 018 that the annotations can be designed for the benefit of the target user, as taught further in par 017 the user’s interests, purchasing history, etc. As this would make reviews more relevant (The teaching above is for the annotations that would be performed in this way), one would be motivated to combine Deluca with Parveen so that a potential user who is reading the review content gets more relevant information. For these reasons, one would be motivated to modify Parveen with Deluca.
Per claims 3 and 13, which are similar in scope, Parveen, Deluca, and Nidell teach the limitations of claims 1 and 11, above. Parveen further teaches wherein the processor is configured to generate the tag content associated with the selected consumer review based on at least one constraint stored in a tag configuration cache in par 0144: “The tag cloud can be made clickable in one of a variety of ways known in the art. For example, an image map can be created based on the random arrangement of features and font sizes. Thereafter, the image map can be laid over the tag cloud in such a manner that clicking on an area of the tag cloud results in clicking an area of the image map corresponding to the clicked area of the tag cloud.” This teaches tag configuration cache because the tag cloud is a tag configuration and it is stored in the image map. The constraint is taught in par 0134 where the sizes of tags are related to their scores.
Per claims 4 and 14, which are similar in scope, Parveen, Deluca, and Nidell teach the limitations of claims 3 and 13, above. Parveen further teaches wherein the processor is further configured to check if the tag content satisfies with the at least one constraint stored in the tag configuration cache, and generate the tag associated with the selected consumer review if the tag content satisfies with the at least one constraint in par 0135: “In some embodiments, the size of the font can instead be used to illustrate the number of reviews with that feature and the color of the font can be used to illustrate the feature score. For example, there can be features with similar font sizes (for example, lens and focus in FIG. 14). In such a case, the color of the words can be used to distinguish which feature has a higher feature score. A legend can be provided to aid a viewer in determining that, for example, the more green a font color is, the more higher the feature score, while the more red a font color is, the lower the feature score is.” If the features have certain numbers of reviews then color of the font ca be used which teaches if the tag content satisfies with at least one constraint, then generate the tag associated….
Per claims 5 and 15, which are similar in scope, Parveen, Deluca, and Nidell teach the limitations of claims 3 and 13, above. Parveen further teaches wherein the processor is configured to generate the tag associated with the selected consumer review further based on search keywords input by a plurality of consumers in par 082: “For example, a user might enter an eCommerce site and type in a search term for a camera. The result could include the layout of FIG. 7 for multiple cameras at once. In such a manner, a user can easily compare different features among different cameras. For example, in FIG. 7, flash has a relatively low score of 37. So some users might view that negatively and decide to purchase a different camera with a higher score for flash. But a different set of users might not be interested in flash performance and might be intrigued by the relatively high body score of 86.” Users teaches search keywords input by a plurality of consumers.
Per claims 8 and 18, which are similar in scope, Parveen, Deluca, and Nidell each the limitations of claims 1 and 11, above. Parveen further teaches wherein the processor is further configured to display the tag associated with the selected consumer review along with information about the service provider included in a list of service providers in pars 062-063: “With reference to FIG. 4, a table 400 is presented that illustrates an exemplary case with multiple reviews and features. Each of columns 410, 420, and 430 represent different reviews. Each of rows 415, 425, 435, and 445 represents different features of the product being reviewed. The intersection of each row and column represents each feature in each review. At the intersection of each row and column is one or more polarities s.sub.i. There can be multiple polarities representing each mention of a particular feature in a particular review. For example, in the review presented above, there can be three different polarities, one for each sentence that mentions a lens.
The table presented in FIG. 4 is merely exemplary. In actual use, there will likely be many more than three reviews and more than four features. In addition, a similar table can exist for each product in a particular database. Some databases contain thousands or even millions of products (for example, the databases of large eCommerce providers).”
Per claims 9 and 19, which are similar in scope, Parveen, Deluca, and Nidell teach the limitations of claims 1 and 11, above. Parveen further teaches wherein the processor is further configured to monitor a user's behaviour for tags displayed on a computing device associated with the user and/or at least one consumer review previously made by the user, and determine which tag of a plurality of tags is to be displayed on the computing device associated with the user based on the monitored information in pars 062-063: “With reference to FIG. 4, a table 400 is presented that illustrates an exemplary case with multiple reviews and features. Each of columns 410, 420, and 430 represent different reviews. Each of rows 415, 425, 435, and 445 represents different features of the product being reviewed. The intersection of each row and column represents each feature in each review. At the intersection of each row and column is one or more polarities s.sub.i. There can be multiple polarities representing each mention of a particular feature in a particular review. For example, in the review presented above, there can be three different polarities, one for each sentence that mentions a lens.
The table presented in FIG. 4 is merely exemplary. In actual use, there will likely be many more than three reviews and more than four features. In addition, a similar table can exist for each product in a particular database. Some databases contain thousands or even millions of products (for example, the databases of large eCommerce providers).”
Per claims 10 and 20, Parveen, Deluca, and Nidell teach the limitations of claims 9 and 19, above. Parveen further teaches wherein the processor is further configured to determine a weight for each of the plurality of tag, based on the monitored information in par 0133: “In this example, tag cloud 1450 includes each of the K features of an exemplary camera that have a feature score over threshold T. As is typical in a tag cloud, tag cloud 1450 displays each of the features with a different font size to differentiate between the features. In this case, the features of tag cloud 1450 are distinguished by the feature score of each feature—the higher the feature score, the larger the font size of the feature.”
Claim(s) 6-7 and 16-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Parveen et al., US PGPUB 20170068999 A1 ("Parveen") in view of Deluca et al., US PGPUB 20190180353 A1 ("Deluca"), further in view of Nidell et al., US PGPUB 20210166282 A1 (“Nidell”), further in view of Chatterjee et al., US PGPUB 20170011029 A1 (“Chatterjee”).
Per claims 6 and 16, which are similar in scope, Parveen, DeLuca and Nidell teach the limitations of claims 1 and 11, above. Parveen further teaches wherein the processor is further configured to determine that the selected consumer review is relevant to two or more categories in par 53: “Thereafter, for every feature in the feature set, the user-generated content (“UGC”) is analyzed to find mentions of each feature (block 304). Thereafter, each of the mentions can be analyzed to determine polarity of each mention (block 306). “ The user generated content teaches consumer reviews. The features are the categories.
See also par 56: :For a feature f.sub.i in the feature set, the number of occurrences in review R.sub.j is n.sub.ij.”
See also par 060. See also Par 050, initial feature set, which as it is determining initial features, teaches predetermined categories (drawn from the specification of the product). See also par 052.
Parveen does not teach and extract two or more phrases each relevant to the two or more categories from the selected consumer review using a natural language processing model.
Chatterjee teaches tagging and scoring techniques for textual passages esp those in social media posts. See abstract.
Chatterjee teaches and extract two or more phrases each relevant to the two or more categories from the selected consumer review using a natural language processing model in pars 0312-0332 which teach various NLP models and in pars 0333-0334 for verbs nouns that are parsed from the reviews which constitute the phrases parsed in par 0335.
It would have been obvious to one ordinarily skilled in the art before the effective filing date of the claimed invention to modify the review annotation and tagging teaching of Parveen with the natural language processing to extract phrases teaching of Chatterjee because Chatterjee teaches in par 09 that: “The combination of machine learning systems with data from human pooled language extraction techniques enables the present system to achieve exceptionally high accuracy of human sentiment measurement and textual categorization of raw text, blog posts, and social media streams. This information can then be aggregated to provide brand and product strength analysis.” As Parveen is determining, in a sense, sentiment, one would be motivated to combine Chatterjee with Parveen because of the taught motivation that exceptionally high accuracy in determining sentiment is taught by Chatterjee. This would improve Parveen’s feature determination and therefore one would be motivated to combine the references.
Per claims 7 and 17, which are similar in scope, Parveen, Deluca, Nidell and Chatterjee teach the limitations of claims 6 and 16, above. Parveen does not teach wherein the processor is further configured to generate two or more tag contents each associated with the two or more phrases.
Chatterjee teaches wherein the processor is further configured to generate two or more tag contents each associated with the two or more phrases in par 0336: “Adaptive rule learning approach will utilize a set of tags that are associated with each sentence and review. These tags are brand, category, polarity, sentiment bearing phrase, category keyword, and vertical. We need to design an efficient and effective User Interface to collect this data quickly and accurately. We will need Multiple Redundant Scoring for brand, category, and polarity. Phrases might vary as well.” See par 0335 for two phrases. See also pars 0337-0352, phrases taught.
It would have been obvious to one ordinarily skilled in the art before the effective filing date of the claimed invention to modify the review annotation and tagging teaching of Parveen with the natural language processing to extract phrases teaching of Chatterjee because Chatterjee teaches in par 09 that: “The combination of machine learning systems with data from human pooled language extraction techniques enables the present system to achieve exceptionally high accuracy of human sentiment measurement and textual categorization of raw text, blog posts, and social media streams. This information can then be aggregated to provide brand and product strength analysis.” As Parveen is determining, in a sense, sentiment, one would be motivated to combine Chatterjee with Parveen because of the taught motivation that exceptionally high accuracy in determining sentiment is taught by Chatterjee. This would improve Parveen’s feature determination and therefore one would be motivated to combine the references.
Therefore, claims 1-20 are rejected under 35 USC 103
Response to Remarks:
35 USC 101
Step 2A: Prong One
Amended independent claims 1 and 11 are not directed to an abstract idea of commercial interactions.
"Software can make non-abstract improvements to computer technology just as hardware
improvements can, and sometimes the improvements can be accomplished through either route. We thus see no reason to conclude that all claims directed to improvements in computer-related technology, including those directed to software, are abstract." Enfish, LLC v. Microsoft Corp., 822 F. 3d 1327. Enfish further explained: "[i]n this case, however, the plain focus of the claims is on an improvement to computer functionality itself, not on economic or other tasks for which a computer is used in its ordinary capacity ... they are directed to a specific improvement to the way computers operate." Further, the USPTO, in its May 19, 2016 Memorandum, acknowledging Enfish, asserts, "[t]he Federal Circuit in Enfish stated that certain claims directed to improvements in computer- related technology, including claims directed to software, are not necessarily abstract (Step 2A) ... the Enfish claims were not ones in which general-purpose computer components are added after the fact to fundamental economic practice or mathematical equation, but were directed to a specific implementation of a solution to a problem in the software art, and concluded that the Enfish claims were directed to an improvement in the operation of a computer technology, thus not directed to an abstract idea (under Step 2A)."
Similar to Enfish, amended independent claim 1 includes computations that are not abstract, but are instead applied in a structured and integrated way that results in an improvement to computer related technology. Independent claim 1 describes a server that generates tag content by summarising selected consumer reviews using a natural language generation engine, generates tags based on that tag content and annotations associated with the selected consumer reviews, and is further configured to classify selected consumer reviews and distribute annotation tasks to computing devices associated with third parties based on that classification. MPEP §2106.04(d)
states that "in a growing body of decisions, the Federal Circuit has distinguished between claims that are 'directed to' a judicial exception (which require further analysis to determine their eligibility) and those that are not (which are therefore patent eligible), e.g., claims that improve the functioning of a computer or other technology or technological field ... and other cases that were eligible as improvements to technology or computer functionality instead of being directed to abstract ideas)." Further, the MPEP 2106.05(a)(II) states that "[t]he courts have also found that improvements in technology beyond computer functionality may demonstrate patent eligibility. In McRO, the Federal Circuit held claimed methods of automatic lip synchronization and facial expression animation using computer-implemented rules to be patent eligible under 35 U.S.C. 101, because they were not directed to an abstract idea." Along similar lines, the focus of claim 1 is on improving how a computer system generates and manages language-based metadata by using a natural language generation engine and classification based task distribution, rather than on any abstract idea in isolation.
The claims cannot be categorized as "certain methods of organizing human activity," because the MPEP defines that category as limited to "fundamental economic principles or practices," "commercial or legal interactions," or "managing personal behavior or relationships or interactions between people" (MPEP §2106.04(a)(2)). Independent claim 1 does not recite economic practices, marketing or sales activities, contractual relationships, or rules governing interactions between people. Rather, the claim recites internal computer operations, including machine-implemented generation of tag content using a natural language generation engine and system-controlled distribution of annotation tasks to computing devices based on classification of digital review data. As explained in MPEP §2106.04(a)(2), the "certain methods of organizing human activity" category is limited to concepts such as fundamental economic principles, commercial or legal interactions, and managing personal behavior or relationships or interactions between people. The claimed generation and management of tag content does not regulate or coordinate human behavior, does not facilitate commerce or legal obligations, and does not organize interpersonal interactions. Instead, it governs how a computer system processes, generates, and routes digital information internally. Accordingly, the claim does not fall within any of the enumerated abstract idea groupings identified by the MPEP.
Examiner responds:
First, the opinion on Enfish does not extend to Applicant’s claims because there are no similarities between the two except both involve computers. Enfish found that claims directed to a self-referential data table for a computer database were not directed to an abstract idea. See MPEP 2106.04(a). Here, the claims are directed to processing consumer reviews and merely apply computer elements to perform this. This is distinct from Enfish because Applicant does not explain why, or provide enough reasoning why, processing consumer reviews is similar to an information subject matter agnostic process such as a self-referential database. Instead, Applicant’s claims are rooted in commonplace routine business processes, which were identified in the rejection. That is why Applicant’s claims are similar to Alice and an abstract idea that is a commercial interaction was properly identified. Tag content, annotations, consumer reviews (consumer – see – business – commercial interaction), are all information elements best described as abstract ideas as they have no form or shape except that they are data and information tied to a commercial interaction. Therefore, this argument is not persuasive.
Applicant argues: The claimed processing apparatus operates in a structured, algorithmic manner within a computer- implemented review-processing architecture to address a concrete technological problem identified in the specification: automatically generating concise, meaningful tag content from unstructured consumer review text at scale while coordinating annotation across multiple external computing devices. The specification explains that manual or conventional approaches to summarizing reviews and generating
tags do not scale with high volumes of consumer reviews and are inconsistent across service providers. To address this problem, the specification discloses that the processor includes a natural language generation engine configured to summarise selected consumer reviews and generate tag content, thereby enabling automated generation of new textual data without manual intervention. The Application further describes that the processor classifies consumer reviews and distributes annotation tasks to third party computing devices based on that classification, thereby enabling distributed, scalable annotation and improving processing efficiency in a multi device computing environment. See as-filed Specification, paragraphs [0062] and [0067]. The claimed features thus improve the functioning of the computer system itself by reducing manual processing, improving scalability, and enabling automated generation and management of language-based metadata.
Therefore, the steps described in amended independent claim 1 constitute a technological
improvement in how computer systems process consumer reviews, through structured classification, task routing, and automated generation of tag content, and are not mere commercial interactions or arrangements of human activity.
Therefore, amended independent claim 1 is not directed to an abstract idea under Prong One of Step 2A.
Applicant argues two improvements. One is that the distribution of tasks to third party computing devices, and the second is that the content is summarized. Per both, an improvement in technology is claimed because something manual, allegedly, is not being done anymore, and now technology is doing it. One ordinarily skilled in the art would not recognize a technical improvement as it is clear that only applied technology is being claimed, not actual technology improved (the “manual process” alleged is replaced by the applied technology). Further it is clear in looking at par 062 that people are the third parties that are performing the task. So in replacing the manual task, the specification describes using manual work ie people performing steps. In par 067 there is summarizing but no technical description as to how this is done and therefore this is an abstract idea. These are argued as done in an “algorithmic manner” which does not support anything except that steps recited were not performed out of order and haphazardly. Therefore these arguments are unpersuasive and the rejection is maintained.
Applicant argues:
Notwithstanding the above remarks, arguendo, that independent claims 1 and 11 are
directed to an abstract idea/judicial exception as the Office Action contends, it is submitted that amended independent claims 1 and 11 recite additional elements that integrate the judicial exception into a practical application.
It is submitted that some steps of amended independent claim 1 are similar to at least claim
1 of Example 42 provided in "Subject Matter Eligibility Examples: Abstract Ideas," issued on January 07, 2019.
Claim 1 of Example 42 recites:
"A method comprising:
a) storing information in a standardized format about a patient's condition in a plurality of network-based non-transitory storage devices having a collection of medical records stored thereon;
b) providing remote access to users over a network so any one of the users can update the information about the patient's condition in the collection of medical records in real time through a graphical user interface, wherein the one of the users provides the updated information in a non- standardized format dependent on the hardware and software platform used by the one of the users;
c) converting, by a content server, the non-standardized updated information into the standardized format,
d) storing the standardized updated information about the patient's condition in the collection of medical records in the standardized format;
e) automatically generating a message containing the updated information about the patient's condition by the content server whenever updated information has been stored; and
f) transmitting the message to all of the users over the computer network in real time, so that each user has immediate access to up-to-date patient information."
The USPTO considers claim 1 of Example 42 to be eligible at Step 2A Prong 2, stating:
"[t]he claim recites a combination of additional elements including storing information, providing remote access over a network, converting updated information that was input by a user in a non- standardized form to a standardized format, automatically generating a message whenever updated information is stored, and transmitting the message to all of the users. The claim as a whole integrates the method of organizing human activity into a practical application. Specifically, the additional elements recite a specific improvement over prior art systems by allowing remote users to share information in real time in a standardized format regardless of the format in which the information was input by the user."
Claim 1 of Example 42 does not merely store information and provide access to the
information but allows remote users to share information in real time in a standardized format regardless of the format in which the information was input by the user. Similarly, amended independent claim 1 meaningfully integrates any alleged abstract idea into a practical application by reciting a coordinated, multi-stage computational architecture that improves the manner in which computer systems process consumer reviews. Claim 1 recites a non-generic technical workflow in which the processor classifies a consumer review, uses that classification as a control signal to distribute an annotation task to a third-party computing device suited to that classification, obtains the returned annotation, generates tag content by summarising the review using a natural language generation engine, and then generates a tag based on both the machine-generated tag content and the externally obtained annotation. These interdependent computer-executed operations form a specific technical arrangement that transforms unstructured review text into system-operational tag metadata and directly governs subsequent system behaviour, thereby improving the computer's9
ability to process, route, and synthesise review-based information. As in Example 42, this represents a practical application of any alleged abstract idea that enhances computer functionality, and therefore satisfies Step 2A Prong 2.
Amended independent claim 1 reflects "an improvement in the functioning of a computer, or an improvement to other technology or technical field." The claim does not merely place known processes on a generic computer to accelerate bookkeeping or display information. Instead, when read as a whole and in light of the specification, the claim recites a purpose-built computer implemented technique for generating language-based metadata and coordinating distributed processing. The natural language generation engine enables the server to autonomously produce new textual artifacts rather than merely extracting or reformatting existing data. By combining automated text generation with classification driven task coordination, the claimed system enhances how computers handle unstructured text, yielding a review processing pipeline that is both scalable and computationally efficient. These features represent a technical advancement over conventional systems that rely on static tagging or manual summarization.
Claim 1 is not directed to commercial interactions, economic relationships, or the
organization of human behavior. Instead, the amended claim defines a computer implemented workflow in which digital review data is processed through model-based generation and system controlled task routing. The apparatus executes this workflow through algorithmic operations that inherently depend on machine processing, including natural language generation and programmatic coordination across external computing devices. These steps do not mirror any human decision making or interpersonal activity, nor do they replicate a business practice implemented on a computer. By requiring model driven generation of tag content and automated distribution logic, the claim establishes a non-conventional processing architecture that addresses computational challenges associated with scale, consistency, and data heterogeneity. The focus of the claim is therefore on improving system behavior, not on structuring human or commercial activity.
Viewed in its entirety, the coordinated architecture of amended claim 1 constitutes a concrete practical application that advances the operation of computer systems in the field of automated review analysis. The claim applies computational tools, for natural language generation, classification, and distributed task coordination, in a manner that produces tangible technical effects: generation of new machine created tag content, efficient allocation of annotation workloads, and improved handling of high volume unstructured text. These effects impose meaningful constraints on any abstract concept by embedding it within a specific server-based implementation that achieves measurable operational benefits. The result is a structured technological process that enhances scalability, reduces manual dependency, and improves consistency in computer
implemented review processing. Accordingly, amended independent claim 1 integrates any alleged abstract idea into a practical application that improves how computer systems generate and manage language-based data, satisfying Step 2A, Prong Two.
Thus, amended independent claim 1 integrates the purported judicial exception into a practical application under Prong Two of Step 2A.
Amended independent claim 11 recites some features that are analogous to the features of amended independent claim 1. Hence, the remarks presented above for amended independent claim
1 apply equally to amended independent claim 11.
Therefore, amended independent claims 1 and 11 recite additional elements that integrate the judicial exception into a practical application under Prong Two of Step 2A.
Examiner responds:
First, Example 42 does not recite an improvement in computer technology. This weighs against finding merit in the previous argument (that there was an improvement to computer technology) because here it is strongly argued that the claims are similar to Example 42. Therefore, if Applicant’s claims are similar enough to Example 42 to warrant overcoming the 101, then Applicant’s claims are not an improvement in computer technology, and the previous argument cannot be maintained. Taken arguendo, Example 42 is not persuasive because unlike these claims that Example had numerous additional elements which, critically, improved over prior art systems (not an improvement in computer technology – see the finding that it was an abstract idea) to allow remote users to share information in real time in a standardized format. In contrast with the instant claims, though, nearly the entirety of Applicant’s claims consist of steps to annotate and summarize reviews with two or three additional elements. One additional element is the generic computer that the steps are run on. This is classically recognized and known by those prosecuting patents as “do it on a computer” and following precedent in Alice is unpatentable. Then, there is the natural language generation engine, which is recited in an applied manner. Taken in combination with the computer this is running an applied algorithm on a computer, where one would naturally find such a thing happening, this combination is still apply it as the algorithm element’s scope is performing in its ordinary capacity, see MPEP 2106.05(f)(2). Then that there is a “computing device associated with a third party” involved is also rejected properly as not a combination like described in Example 42. This is because as shown in that MPEP section cited a “mobile unit and server” working together is apply it. There is no analogy to the specific additional elements and improvement statement found in Example 42. Instead, these claims have a lack of additional elements and no plausible improvement explanation recited in the specification (Par 062 establishes, pretty plainly, that the improvement is assigning tasks to people to have them do them, not automated in the same way that a rideshare service may arrange a driver but the driver still drives the car). Finally, there are critical elements in Ex 42 that have no analog here. One is the standardized format, therefore the claims are limited by a format and this is not 12 pt Arial but something computers see, like .doc .docx .pdf .xlsx etc. The other are timeliness elements, automatically generating a message “whenever updated information is stored” and then transmitting in real time to all of the users (plural) over a network. Therefore, this has been carefully considered, but Example 42 is not similar to Applicant’s claims except both have computers and both involve multiple parties (people).
Applicant argues:
Notwithstanding the above remarks under Step 2A, arguendo, that independent claims 1 and 11 are directed to an abstract idea/judicial exception as the Office Action contends, it is submitted that amended independent claims 1 and 11 amount to "significantly more" than an abstract idea.
According to the MPEP, "[e]valuating additional elements to determine whether they
amount to an inventive concept requires considering them both individually and in combination to ensure that they amount to significantly more than the judicial exception itself ... Consideration of the elements in combination is particularly important, because even if an additional element does not amount to significantly more on its own, it can still amount to significantly more when considered in combination with the other elements of the claim" (emphasis added). See MPEP 2106.05(I).
Reconsideration and withdrawal of the rejection of independent claim 1 under 35 U.S.C. § 101 are requested in view of the decision of the Court of Appeals for the Federal Circuit regarding subject matter eligibility as acknowledged by the USPTO in its Memorandum issued on November 2, 2016, addressing "Recent Subject Matter Eligibility Decisions (BASCOM Global Internet Services v. A T&T Mobility LLC)." The USPTO, in the Memorandum, instructs "[t]he BASCOM court agreed that the additional elements were generic computer, network, and Internet components that did not amount to significantly more when considered individually, but explained that the district court erred by failing to recognize that when combined, an inventive concept may be found in the non-conventional and non-generic arrangement of the additional elements ... (note that the term 'inventive concept' is often used by the courts to describe additional element(s) that amount to11
significantly more than a judicial exception)" (emphasis in original). The USPTO, in the Memorandum, further describes "[i]n Step 2B of the USPTO's SME guidance, examiners should consider the additional elements in combination, as well as individually, when determining whether a claim as a whole amounts to significantly more, as this may be found in the non-conventional and non-generic arrangement of known, conventional elements."
Amended independent claim 1 amounts to "significantly more" at least because the
"additional elements" add "a specific limitation other than what is well-understood, routine and conventional in the field, or adding unconventional steps that confine the claim to a particular useful application."
The Application identifies limitations in conventional approaches for processing consumer reviews, which primarily rely on centralized, system internal processing and predetermined rules. In particular, the specification explains that prior approaches depend on predetermined or fixed rules that do not adapt over time, requiring subsequent updating based on received annotation input. Such static processing limits accuracy and effectiveness, especially as the nature and diversity of consumer reviews evolve. The specification further highlights that purely internal automated analysis is insufficient for accurately annotating heterogeneous review content across different categories and languages, prompting the use of external annotation expertise. Collectively, these passages identify conventional practices as centralized, static, and non-adaptive, with limited scalability and reduced annotation accuracy. See Application paragraph [0004]-[0007], [0062], and [0071].
In contrast, amended independent claim 1 describes non-conventional computational steps that involve classifying a selected consumer review and using that classification as a control signal to distribute an annotation task to computing devices associated with third parties, followed by generating tag content using a natural language generation engine and generating tags based on both the machine generated tag content and the externally obtained annotation. This provides a structured, machine executed workflow in which classification, task distribution, third party annotation, and tag generation operate as interdependent stages of a single server-controlled process. Thus, these operations go beyond mere commercial practices or abstract organization of information, as they require coordinated, interrelated computational steps executed by a processor and memory architecture that culminate in automated system level generation and use of tags and annotations. See as-field Specification, paragraphs [0062] and [0067].
The Application introduces automated summarization through a natural language generation engine as a response to the absence of mechanisms in prior approaches for generating concise, synthesized tag content from review text, rather than relying on extracted keywords or manually
curated tags. As emphasized in BASCOM, the claimed system thus amounts to "significantly more" than an abstract idea because it implements a specific technical architecture - a non-conventional arrangement of otherwise known components, that improves computer functionality and enhances the technical field of computer implemented review processing by enabling scalable, distributed annotation and automated generation of structured tag content from unstructured consumer review data.
In view of the foregoing remarks, amended independent claim 1, when taken as a whole, qualifies as significantly more than an abstract idea.
Amended independent claim 11 recites some features that are analogous to the features recited by amended independent claim 1. Hence, the remarks presented above for amended independent claim 1 apply equally to amended independent claim 11.
Therefore, independent claims 1 and 11 amount to "significantly more" than an abstract idea under Step 2B.
Examiner replies:
Applicant’s arguments are not to the rejection which found that the elements in combination were “apply it” and there is no finding of the conventionality or unconventionality of the elements. Therefore Applicant’s arguments do not overcome the finding of “apply it” in step 2B. As shown above in the rejection and here in the MPEP, the findings of Prong 2A step 2 can be carried over into step 2B. As Applicant has not overcome the finding that the combination of additional elements are apply it, Applicant has not overcome the step 2B element of the rejection, that apply it elements are not significantly more than the abstract idea.
See MPEP 2106.05:
Although the conclusion of whether a claim is eligible at Step 2B requires that all relevant considerations be evaluated, most of these considerations were already evaluated in Step 2A Prong Two. Thus, in Step 2B, examiners should:
• Carry over their identification of the additional element(s) in the claim from Step 2A Prong Two;
• Carry over their conclusions from Step 2A Prong Two on the considerations discussed in MPEP §§ 2106.05(a) - (c), (e) (f) and (h):
Applicant’s arguments have been reviewed and carefully considered, but they do not address the actual rejection.
35 USC 103
Parveen, at best, describes that all feature identification, classification, annotation, and
display operations are performed internally by the same system. However, Parveen fails to teach or suggest classifying a selected consumer review for the purpose of distributing an annotation task to a computing device associated with a third party, as required by independent claim 1. Parveen also fails to teach obtaining annotations from third party computing devices or generating tags based on a combination of machine generated tag content and externally obtained annotations.
Examiner responds:
This is unpersuasive as it attempts to narrow the reference based on the actual equivalent terms which Parveen teaches the claimed limitations. This also does not address the particulars of the rejection.
Applicant argues:
Parveen, in paragraph [0114], describes feature "tags" as visual indicators derived from
internally calculated feature scores. Further, paragraphs [0116] and [0118] of Parveen describe user interaction with those tags after annotation is complete, such as filtering or viewing reviews associated with a selected feature. Paragraph [0120] of Parveen describes graphical summaries that visualize sentiment already computed by the system. None of these descriptions describe assigning annotation work, distributing annotation tasks, or routing annotation responsibilities to other computing devices. Further, in Parveen, feature classification and annotation are outputs of internal analysis used to organize and display review information. There is no disclosure that classification is used as a control input that governs where annotation work is performed. There is also no disclosure of third party computing devices performing annotation in response to distributed tasks.
Examiner responds: Applicant’s argument not to the rejection, the rejection is:
Parveen teaches generate a tag content associated with the selected consumer review by summarizing the selected consumer review in par 114: “Column 1020 contains a list of “tags” that represent features that received either positive or negative feature scores in each review. In some embodiments, the tags can be color-coded. In such a manner, one color can represent positive feature scores and another color can represent negative feature scores. Instead of color coding, some embodiments can use other methods of indicating the difference between positive, negative, and neutral feature scores, such as shading, hatch marks, underlining, typeface, font colors, and the like.” See for generated par 0107: “And the features with a feature score less than a given threshold score (e.g., zero), can get marked as “Bad.” Then the reviews can be displayed to the user with a tag indicating the feature and whether the feature score was “Good” or “Bad” (block 910). Displaying a review to the User can involve transmitting data to a user's computer that cause the user's computer to display reviews.”
If “third party computing devices” are a desired a positively claimed element, then they should be recited in the independent claims. Examiner believes that the element, or really a description (“associated with a third party”) is being extended to “third party computing devices.” Perhaps not, it’s not clear. At any rate, it would be necessary to amend to say “third party computing device” which is nowhere in the claims. Here there is a “computing device associated with a third party” but DeLuca is asserted to teach that and not Praveen, so the argument would not be persuasive given it is not related to the rejection.
Here Parveen is generating tag content which has a broad scope and all it has to do is summarize the selected consumer review which could be by a color, for example. As the remarks don’t address this the remarks are therefore unpersuasive.
Applicant argues:
In contrast, the Application expressly states that "the processor 120 may distribute the task for the annotation... to computing devices associated with one or more third parties based on the classification of the selected consumer review." The Application further explains that "third parties... subscribe categories and languages of consumer reviews they are comfortable to give accurate annotation," such that task assignment is driven by review classification. See as-filed Specification, paragraph [0062]. Parveen does not disclose or suggest this architecture and therefore does not render independent claim 1 obvious.
Examiner responds: “third party computing device” is not claimed, but “computing device associated with a third party” is claimed, they are not equivalent in scope as a computing device can be associated with a third party and also other parties, (what is claimed) but a third party computing device is more likely just a computing device for the third party.
Also Parveen is not asserted to teach this final limitation, which is the distribute a task limitation, so this argument does not pertain to the rejection.
Applicant argues:
Further, the Office Action concedes that "Parveen does not teach and distribute a task for the annotation associated with the selected consumer review to the computing device associated with the third party based on the classification of the selected consumer review." The Office Action then relies on Deluca to teach this feature.
However, Deluca describes a vendor management computer system that normalizes and presents product reviews within a virtual storefront based on characteristics of a user or a target user. The system collects information about a user or target user and applies that information to prioritize, reorder, or filter reviews presented in the storefront. The system further annotates normalized reviews to highlight portions that it determines to be relevant to the target user and generates summaries explaining why certain reviews or sections are emphasized. Internal system modules generate these annotations and summaries as part of a presentation and navigation workflow within the storefront interface. See Deluca, paragraphs [0016]-[0018], [0056]-[0060], and [0069].
Deluca, at best, describes internally annotating and summarizing reviews to improve
presentation and relevance for a target user within the virtual storefront interface. However, Deluca fails to teach or suggest distributing annotation tasks to computing devices associated with third parties based on classification of consumer reviews, as required by amended independent claim 1. Deluca also fails to teach obtaining annotations from third party computing devices or generating tags based on a combination of externally obtained annotations and machine generated tag content.
Deluca in paragraph [0056] describes collecting user or target user data to normalize and
prioritize the presentation of reviews. Further in paragraph [0069] Deluca explains that reviews may be annotated and summarized to explain relevance to a user. In each case, the annotation and summarization are clearly disclosed as functions of an internal vendor management system module operating to customize review display. Deluca neither describes distributing annotation tasks outside the system, nor does it describe third party computing devices performing annotation work and using classification of reviews as a mechanism to route annotation tasks. Instead, classification and analysis in Deluca are used solely to determine ordering, highlighting, and presentation of review content.
By contrast, the Application discloses annotation performed by third party computing devices, where tasks are distributed based on classification of the consumer review. The Application further discloses that tag generation relies on machine generated tag content produced using a natural language generation engine, which is then combined with externally obtained annotations to generate tags. See as-filed Specification, paragraphs [0062] and [0067]. Deluca contains no such teaching or suggestion of this claimed workflow.
Examiner Responds:
Applicant cites par 056 but here is where a user is providing input about a product review or information relating to a product review. Here a task has been distributed “for the annotation” not necessarily the annotation itself but by normalizing the review this is “for the annotation” of the review. While, on the client device, it is performed on a “computer associated with a third party” (item 111), it is also performed by being displayed on the “User’s client device” par 060. It is possible that the scope of “task for the annotation” was intended to be solely “annotation” which might distinguish from Deluca however that is not what was claimed.
Per the natural language engine, DeLuca is not asserted to teach that and that argument is moot per further search and consideration.
Applicant argues:
Even in combination, the cited references fail to teach or suggest the above-mentioned features of amended independent claim 1. Parveen's feature tags and annotations are internally generated markers used for review navigation and display, while Deluca's annotations and summaries are internally generated to customize presentation for a target user. Neither reference teaches or suggests distributing annotation tasks to third party computing devices, nor do they teach using classification of a consumer review as a control mechanism for such distribution. Further, neither reference discloses generating tags based on a fusion of machine generated tag content and externally obtained annotations, as expressly set forth in the claimed Application. Any conclusion to the contrary would require impermissible hindsight reconstruction using the claimed Application as a roadmap
Therefore, in view of the above, Parveen and Deluca, either alone or in combination, fail to teach or suggest the above-mentioned claim features of amended independent claim 1.
Independent claim 11 recites some or all subject matter similar to independent claim 1.
Therefore, all the remarks made for independent claim 1 above, apply equally to independent claim 11.
Therefore, the rejection of independent claims 1 and 11 under 35 U.S.C. § 103 should be withdrawn.
Dependent claims 2-5, 8-10, 12-15, and 18-20
Examiner responds: Examiner shows the construction above as teaching the claimed elements and clarifies in arguments where Praveen and Deluca teach the limitations that were argued were not taught. The conclusion reached that there must be impermissible hindsight is not justified as there is a teaching suggestion motivation that was found that one ordinarily skilled in the art would use to combine the references.
Applicant argues:
Parveen merely describes internally analyzing consumer reviews to identify features and calculate feature scores used to annotate and display reviews. The feature scores are derived from sentiment polarity and are used to mark annotations as positive, negative, or neutral for display purposes. The annotations and scores are generated entirely within the system based on automated sentiment analysis. However, Parveen fails to teach or suggest updating a rule for selecting relevant consumer reviews based on an annotation obtained from a computing device associated with a third party, as required by claim 2.
Indeed, in Parveen any "rules" applied are static analytical procedures for calculating sentiment polarity and aggregating feature scores. Parveen describes that feature scores are calculated by analyzing mentions and polarities in review text and then displayed or labeled accordingly. There is no disclosure that annotations are supplied by third party computing devices, nor that such annotations are used as feedback to adapt or update review selection logic. All processing in Parveen is centralized and system internal. By contrast, claim 2 requires that the processor updates a rule for selecting relevant consumer reviews based on externally obtained annotations, which Parveen never discloses or contemplates.
Therefore, Parveen fails to teach or suggest the features of dependent claim 2.
Deluca merely describes normalizing and prioritizing the presentation of product reviews based on characteristics of a user or target user. The user or target user data is collected and analyzed to reorder, highlight, or annotate reviews for display. The annotations are generated by internal system modules and are used to explain relevance to a user. However, Deluca does not disclose updating the underlying rules used to select which reviews are relevant based on annotation feedback.
Indeed, in Deluca, normalization logic is driven by stored user or target user characteristics and analytics applied by the vendor management system. While Deluca describes annotating and summarizing reviews, those annotations are presentation oriented and generated internally to explain relevance. There is no disclosure that annotations are obtained from third party computing devices, nor that any such annotations are used to update a rule governing which consumer reviews are selected as relevant.
Therefore, Deluca fails to teach or suggest the features of dependent claim 2.
Combination: Even if Parveen and Deluca are combined, the resulting system would still fail to teach or suggest the features of claim 2. Parveen's system performs internal sentiment analysis using fixed analytical procedures, while Deluca's system applies stored user profile characteristics to normalize review presentation. Neither reference teaches or suggests using annotations obtained from third party computing devices as feedback to update a rule for selecting relevant consumer reviews as opposed to the Application, which expressly describes rule updating based on third party annotation feedback.
Therefore, Parveen and Deluca, either alone or in combination, fail to teach or suggest the features of dependent claim 2.
Examiner responds:
DeLuca teaches this limitation by showing that the reviews presented (selected) are done so in response to changes put in that describe the profile of the user viewing the reviews. This is based on the same “annotation feedback” (see above discussion where these arguments diverge from the claim language) as was asserted in claim 1 and therefore this rejection is proper. The arguments do not adhere to the claim language; once the claim language is analyzed, the rejection is shown to be correct. If the claim language is desired to fit the arguments made (“annotation feedback” found nowhere in the claims) then the claim language must be changed. Therefore the rejection is maintained.
Applicant argues:
Chatterjee is not contended to cure the deficiencies in the rejection of independent claim 1 above.
Nevertheless, Chatterjee describes a hybrid human-machine learning system in which documents are divided into "pieces" of text that are tagged and scored using a combination of machine-learning engines and human crowd-workers. The system includes a crowdsourcing module that selects pieces of text and presents them to humans for annotation with attributes such as sentiment magnitude, category, or topic. Human-generated labels and machine-generated labels are then aggregated, and the aggregated data is fed into a machine-learning module to train or update classification models. The sentiment analysis engine further parses text into multiple "pieces" and performs piece-level scoring and categorization. See Chatterjee, paragraphs [0009]-[0011], [0032], [0039]-[0041], [0047]-[0051], and [0098]-[0104].
Chatterjee merely describes that the hybrid human-machine learning system divides text documents into "pieces," and those pieces are scored or tagged for sentiment and category information. However, Chatterjee fails to teach or suggest determining that a single consumer review is relevant to two or more predetermined categories and extracting, from within that review, two or more phrases each specifically relevant to the two or more categories using a natural language processing model, as required by claim 6.
Although Chatterjee discusses "pieces" of text and category assignment, those pieces are
defined as units for scoring and training purposes, typically derived through human labeling or rule based segmentation. Chatterjee does not describe an NLP driven process that automatically identifies a single consumer review as relevant to multiple predefined categories and then extracts separate phrases corresponding to those categories as part of system controlled review processing. Instead, Chatterjee relies on crowd sourced tagging of attributes and subsequent aggregation for learning and analysis. Thus, Chatterjee fails to teach or suggest the features of claim 6.
Parveen merely describes identifying features discussed in consumer reviews and associating those features with sentiment information derived from automated analysis. In Parveen, a review
may mention multiple features, and sentiment polarity is calculated for each identified feature and aggregated into feature scores for display purposes. However, Parveen fails to teach or suggest determining that a single consumer review is relevant to two or more predetermined categories and extracting two or more phrases each specifically relevant to the two or more categories from the review using a natural language processing model, as required by claim 6.
Indeed, Parveen aggregates sentiment signals associated with detected features and presents summary indicators or tags for navigation and display. Claim 6, by contrast, requires a structured NLP driven operation that decomposes a review into multiple category relevant phrases, a processing step absent from Parveen.
Therefore, Parveen fails to teach or suggest the features of dependent claim 6.
Deluca merely describes parsing consumer reviews to identify portions of text relevant to a target user and annotating or highlighting those portions to improve presentation and relevance. Deluca uses user or target user characteristics to determine which reviews or portions of reviews should be emphasized. However, Deluca fails to teach or suggest determining that a single consumer review is relevant to two or more predetermined categories and extracting separate phrases corresponding to those categories using a natural language processing model, as required by claim 6.
Therefore, Deluca fails to teach or suggest the features of dependent claim 6.
Examiner responds:
The arguments, similar to before, do not follow the claim language. Chatterjee teaches the limitations by finding features in the text (whether they are consumer reviews or not is of no import as that element is already taught by the first to limitations and consumer reviews are taught by blocks of text) using natural language processing. Therefore Chatterjee will be maintained.
Applicant argues:
Even when Parveen, Deluca, and Chatterjee are considered together, they fail to render
claim 6 obvious. Parveen identifies features and aggregates sentiment, Deluca highlights or annotates review portions for personalization, and Chatterjee assigns categories and sentiment scores to text pieces for scoring and learning. None of the references, alone or in combination, teach or suggest a workflow in which a processor (i) determines that a single consumer review is relevant to two or more predetermined categories and (ii) extracts two or more phrases each specifically relevant to those categories using a natural language processing model.
Therefore, Parveen, Deluca, and Chatterjee, either alone or in combination, fail to teach or suggest the features of dependent claim 6.
As a result, the rejection of dependent claims 6 and 16 under 35 U.S.C. § 103 should be withdrawn.
Dependent claims 7 and 17
Chatterjee is not contended to cure the deficiencies in the rejection of independent claim 1 above. Accordingly, claims 7 and 17 are also allowable through their respective dependency on amended independent claims 1 and 11.
As a result, the rejection of dependent claims 7 and 17 under 35 U.S.C. § 103 should be withdrawn.
Examiner responds:
As was found in the statement of motivation, there is a teaching suggestion or motivation above that was asserted and is not argued against here. Therefore the argument, while carefully considered, does not address the actual rejection and the rejection is maintained.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RICHARD W. CRANDALL whose telephone number is (313)446-6562. The examiner can normally be reached M - F, 8:00 AM - 5:00 PM.
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, Anita Coupe can be reached at (571) 270-3614. 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.
/RICHARD W. CRANDALL/ Primary Examiner, Art Unit 3619
Prosecution Insights