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 .
Response to Amendment
This Office action is in response to Applicant's amendment filed on 12/5/2025.
Claim 1-10, 12, 14-20 are pending. Claim 1 and 14 are amended. Claim 11 and 13 are cancelled. Claim 1-10, 12, 14-20 are rejected.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim 1, 3-6, 9-10 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Chauhan, Sejal et al (PGPUB Document No. 20220067052), hereafter referred as to “Chauhan”, in view of Thonet, Thibaut et al (US Patent No. 12399945), hereafter, referred to as “Thonet”, in view of Kanno, Kyota (WIPO Publication WO2011049037), hereafter, referred to as “Kanno”, in further view of Shemesh, Ariel et al (PGPUB Document No. 20200005119), hereafter, referred to as “Shemesh”.
Regarding Claim 1(Currently Amended), Chauhan teaches A method, comprising: creating a generic embedding for a query(Chauhan, para 0089-0090 disclose a process for creating/establishing a generic/default embedding or vector for users for whom no user data is available “a default vector 304 may be established …. The default vector 304 and items 302 may be input to search engine 106 such that a default list of items 313 may be presented through user interface 312. For example, such a list may be appropriate to certain types of users, e.g., new users (for whom no user data is available), basic users” ); receiving, at a user-specific morph operator, the generic embedding for a query as input, wherein the user-specific morph operator encodes user preferences for a user into a matrix; transforming, using the user-specific morph operator, the generic embedding for the query to a personalized embedding for the query that-- personalizes the query to the user preferences(Chauhan, para 0042 discloses a default or generic vectors/embeddings/matrix transforming into personalized embedding by applying users’ identified preferences (morph operator)“a default vector may be preserved for use in general searches, administration, default rankings for items, and other uses, while periodically updated clusters of users and corresponding machine learning models may be used to generate cluster-specific vectors to rerank search results such that users in a particular cluster are provided ranked search results that are likely closer to the preferences of the users that belong in the particular cluster”),
and provide, in response to the querying, a recommendation with one or more items to present for the query that are related to the query and incorporate preferences of the user based on the personalized embedding(Chauhan, para 0093 discloses providing a search result based on modified/personalized embeddings which reflects user’s personal preferences “The search engine 106 may further provide a customized ranked list 315 and 317 based on the modified vectors 308 and 310, respectively. Each customized ranked list 315 and 317 may be presented to a user…. have a customized ranked list 315 presented in response to a search query that is customized based on the user data 201 and associated profile data and behavioral data” ).
Using the broadest reasonable interpretation consistent with the specification (paragraph 0030) as it would be interpreted by one of ordinary skill in the art, examiner is interpreting the limitation “user-specific morph operator” to mean identified user preference features which dictates user preferences.
Chauhan does not explicitly teach wherein creating the personalized embedding for the query provides personalization on the query and items retrieved while keeping a corpus storing the items non personalized providing two-sided personalization on the query and item side while incurring a cost of one-sided personalization by using the matrix decomposition of the user-specific morph operator into multiple matrices; querying, using the personalized embedding for the query, a maximum inner product search (MIPS) structure that modifies retrievals for the query personalized to the user;
However, in the same field of endeavor of query embeddings Thonet teaches wherein creating the personalized embedding for the query provides personalization on the query and items retrieved while keeping a corpus storing the items non personalized providing two-sided personalization on the query and item side while incurring a cost of one-sided personalization by using the matrix decomposition of the user-specific morph operator into multiple matrices(Thonet, claim 1 discloses providing two-sided personalization by updating the embedding first with user query and second with item “receiving training data including search instances and recommendation instances, each of the search instances including (a) a user profile, (b) a query comprising one or more query terms, and (c) an item………….. computing, based on the base embeddings and the respective embeddings obtained from each of the one or more convolutional layers: a first loss determined based on a prediction by the HGCN of an item based on a corresponding user profile and query; and a second loss determined based on a prediction by the HGCN of a query based on a corresponding user profile and item; and selectively updating at least one of the base embeddings of the HGCN based on the first loss and the second loss”; where items are being used for updating the base embedding for personalization but no update to items are being performed; where the limitation “by using the matrix decomposition of the user-specific morph operator into multiple matrices” is taught by prior art Kanno discussed later);
Using the broadest reasonable interpretation consistent with the specification (paragraph 0028 and 0033) as it would be interpreted by one of ordinary skill in the art, examiner is interpreting the limitation “providing two-sided personalization on the query and item side while incurring a cost of one-sided personalization” to mean
having both query and item embeddings in the same embedding space that represents user personalization embedding to achieve single cost of personalization by avoiding two separate embedding.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of updating basic embedding for personalization of Thonet into the feature of generic and personal embeddings of Chauhan to produce an expected result of making embeddings personalized with query/intent features and its corresponding item features. The modification would be obvious because one of ordinary skill in the art would be motivated to determine the deficiencies (loss) in personalization in embeddings and updating the personalized embeddings accordingly (Thonet, abstract).
But Chauhan and Thonet don’t explicitly teach by using the matrix decomposition of the user-specific morph operator into multiple matrices; querying, using the personalized embedding for the query, a maximum inner product search (MIPS) structure that modifies retrievals for the query personalized to the user;
However, in the same field of endeavor of query embeddings for personalization Kanno teaches by using the matrix decomposition of the user-specific morph operator into multiple matrices(Kanno, claim 6 (page 12 line 31-36) discloses decomposing user specific matrix (user-specific morph operator) into plurality of matrices for content personalization “A preference matrix in which user preference information for each item of each user is arranged in n rows and m columns, where n is the number of users and m is the number of items, is n rows and k columns (k is an integer smaller than m with 2 or more). A preference concept vector extraction unit that decomposes into a user's preference concept matrix and a preference concept matrix of items of k rows and m columns”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of decomposing user preference matrix of Kanno into the feature of item and query embeddings of Chauhan and Thonet to produce an expected result of embedding personalization. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the item recommendation by personalization(Kanno, claim 6).
But Chauhan, Thonet and Kanno don’t explicitly teach querying, using the personalized embedding for the query, a maximum inner product search (MIPS) structure that modifies retrievals for the query personalized to the user;
However, in the same field of endeavor of query embeddings Shemesh teaches and querying, using the personalized embedding for the query, a maximum inner product search (MIPS) structure that modifies retrievals for the query personalized to the user(Shemesh, para 0065 discloses that querying using user vector/embeddings (personalized) and MIPS which performs search operations and obtain products which are similar by measures “Inner product search, or maximum inner product search, as used herein, refers to a search that, for a given user vector, performs a dot product of the user vector with all the candidate vectors, and selects a number of candidate vectors (e.g., top K out of all N candidate vectors) with maximum dot product. In this way, the inner product search relates to nearest neighbor search, which finds the point in a given set that is closest (or mostly similar) to a given point, only that, in this specific case, the closeness or similarity is defined in terms of an inner product similarity measure” ).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of using MIPS of Shemesh into the feature of generic and personal embeddings of Chauhan, Thonet and Kanno to produce an expected result of making embeddings personalized in real time. The modification would be obvious because one of ordinary skill in the art would be motivated to reduce the computational cost by filter-wise implementing MIPS(Shemesh, abstract).
Regarding claim 3(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 and further teaches wherein the query is an interest of the user(Chauhan, para 0093 discloses personalized query which implies queries are related to user interest “a user assigned to cluster C1 may have a customized ranked list 315 presented in response to a search query that is customized based on the user data 201 and associated profile data and behavioral data”).
Regarding claim 4(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 and further teaches wherein the interest of the user is inferred from multiple event interactions by the user(Chauhan, para 0083 discloses personalized query related to user interest are drawn from user interaction events such as item purchases, view etc. “behavioral data features 205 can include purchase history (e.g., bought at least one item last week), types of items purchased, types of items viewed, gameplay history (e.g., played “pizza place” daily for a month), types of games played”).
Regarding claim 5(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 and further teaches wherein the query is any user representation(Chauhan, para 0090 discloses generic or default embedding is being used for representing any user before personalization “The default vector 304 and items 302 may be input to search engine 106 such that a default list of items 313 may be presented through user interface 312. For example, such a list may be appropriate to certain types of users, e.g., new users (for whom no user data is available)”).
Regarding claim 6(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 and further teaches wherein the personalized embedding for the query retains information for the query and incorporates the user preferences for the query(Chauhan, para 0093 discloses incorporating/adding user preferences/personalized embedded features to query for providing personalized contents “The search engine 106 may further provide a customized ranked list 315 and 317 based on the modified vectors 308 and 310, respectively. Each customized ranked list 315 and 317 may be presented to a user…. have a customized ranked list 315 presented in response to a search query that is customized based on the user data 201 and associated profile data and behavioral data”).
Regarding claim 9(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 and further teaches wherein the user-specific morph operator is automatically learned using a user history of previous actions performed by the user or using user profile information(Chauhan, para 0088 discloses user specific features are learned from user profile using machine learning in a neural network “The associated machine learning model may be trained (either supervised or unsupervised) to create modified feature vectors for items that facilitate customized rankings of items for search results and/or listings through the search engine 106”; where para 0080-0083 further teaches profile or user historical data is being used for updating model “The user data 201 may include, but is not limited to, User IDs, profile data features 203, behavioral data features 205 (behavioral parameters), and/or other data features….”).
Regarding claim 10(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 and further teaches wherein applying the user-specific morph operator to the generic embedding further comprises: incorporating the user preferences into information for the query(Chauhan, para 0093 discloses incorporating/adding user preferences/personalized embedded features to query for providing personalized contents “The search engine 106 may further provide a customized ranked list 315 and 317 based on the modified vectors 308 and 310, respectively. Each customized ranked list 315 and 317 may be presented to a user…. have a customized ranked list 315 presented in response to a search query that is customized based on the user data 201 and associated profile data and behavioral data”).
Claim 11, Cancelled.
Regarding claim 12(Previously Presented), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 and further teaches wherein the one or more items are selected based on the user preference for the user (Chauhan, para 0093 discloses recommending/selecting contents based on user preferences “a customized ranked list 315 presented in response to a search query that is customized based on the user data 201 and associated profile data and behavioral data”).
Claim 13, Cancelled.
Claim 2 is rejected under 35 U.S.C. 103 as being unpatentable over Chauhan, Sejal et al (PGPUB Document No. 20220067052), hereafter referred as to “Chauhan”, in view of Thonet, Thibaut et al (US Patent No. 12399945), hereafter, referred to as “Thonet”, in view of Kanno, Kyota (WIPO Publication WO2011049037), hereafter, referred to as “Kanno”, in view of Shemesh, Ariel et al (PGPUB Document No. 20200005119), hereafter, referred to as “Shemesh”, in further view of Hsieh, Cheng-Kang et al (PGPUB Document No. 20210174164), hereafter, referred to as “Hsieh”.
Regarding claim 2(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 but don’t explicitly teach wherein the query is an event performed by the user.
However, in the same field of endeavor of query embeddings Hsieh teaches wherein the query is an event performed by the user(Hsieh, para 0122 discloses that query is triggered by an event associated with a user “The user data may be processed ahead of time or on demand in response to a triggering event (such as the receipt of a query input associated with a user)” ).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of triggering queries by events of Hsieh into the feature of generic and personal embeddings of Chauhan, Thonet, Kanno and Shemesh to produce an expected result of making embeddings personalized in real time. The modification would be obvious because one of ordinary skill in the art would be motivated to use content-centric predictive personalization to recommend contents to new users in real or run time where no prior data for user personalization is available(Hsieh, para 0030).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Chauhan, Sejal et al (PGPUB Document No. 20220067052), hereafter referred as to “Chauhan”, in view of Thonet, Thibaut et al (US Patent No. 12399945), hereafter, referred to as “Thonet”, in view of Kanno, Kyota (WIPO Publication WO2011049037), hereafter, referred to as “Kanno”, in view of Shemesh, Ariel et al (PGPUB Document No. 20200005119), hereafter, referred to as “Shemesh”, in further view of Shukla, Anand et al (US Patent No. 11294974), hereafter, referred to as “Shukla”.
Regarding claim 7 (Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 but does not explicitly teach wherein the generic embedding for the query is in response to the user performing an event.
However in the same field of endeavor of query embeddings Shukla teaches wherein the generic embedding for the query is in response to the user performing an event (Shukla, where claim 15 teaches embedding query terms (or generic embedding) for querying “generating an embedding for the one or more query terms; determining one or more web documents that are similar to the generated embedding”; where col 5:30-35 further discloses query embedding can be created on-the-fly in response to a query event “the search and feed service may generate an embedding for the long-tail entity in response to the query (e.g., “on the fly”). To generate the embedding for the long-tail entity, the search and feed service may perform a search based on the one or more query terms and the search is configured to return a plurality of web documents”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of generating embeddings for query event of Shukla into generic and personalized embeddings of Chauhan, Thonet, Kanno and Shemesh to produce an expected result of making embeddings on they fly. The modification would be obvious because one of ordinary skill in the art would be motivated to reduce the number of computations by having pre-computed embeddings for a limited number of entities (Shukla, col 4:55-60).
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Chauhan, Sejal et al (PGPUB Document No. 20220067052), hereafter referred as to “Chauhan”, in view of Thonet, Thibaut et al (US Patent No. 12399945), hereafter, referred to as “Thonet”, in view of Kanno, Kyota (WIPO Publication WO2011049037), hereafter, referred to as “Kanno”, in view of Shemesh, Ariel et al (PGPUB Document No. 20200005119), hereafter, referred to as “Shemesh”, in further view of Zhang, Chi et al (PGPUB Document No. 20230245418), hereafter, referred to as “Zhang”.
Regarding claim 8(Original), Chauhan, Thonet, Kanno and Shemesh teach all the limitations of claim 1 but does not explicitly teach wherein the generic embedding represents the query by encoding a textual description of the query using a machine learning model.
However in the same field of endeavor of query embeddings Zhang teaches wherein the generic embedding represents the query by encoding a textual description of the query using a machine learning model (Zhang, para 0067 teaches extracting query texts and generic or general query specific embeddings using extracted query text “The query text embedding module 122 is configured to, upon receiving the extracted query text feature, embed the query text feature to query text embeddings, and send the query text embeddings to the query transformer 128”; where para 0062 discloses feeding query text features to machine learning model for encoding “In certain embodiments, the concatenation sequence between text feature and image feature is fed into a machine learning model, such as a transformer encoder model”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of query embeddings of Zhang into generic and personalized embeddings of Chauhan, Thonet, Kanno and Shemesh to produce an expected result of making embeddings on they fly. The modification would be obvious because one of ordinary skill in the art would be motivated to enhance the search process by providing a multi-modal searching options to users (text to text, image to image etc.) (Zhang, para 0063 & 100).
Claim 14-15 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Chauhan, Sejal et al (PGPUB Document No. 20220067052), hereafter referred as to “Chauhan”, in view of LuVogt, Chris et al (PGPUB Document No. 20130290110), hereafter, referred to as “LuVogt”, in view of Thonet, Thibaut et al (US Patent No. 12399945), hereafter, referred to as “Thonet”, in view of Kanno, Kyota (WIPO Publication WO2011049037), hereafter, referred to as “Kanno”, in further view of Shemesh, Ariel et al (PGPUB Document No. 20200005119), hereafter, referred to as “Shemesh”.
Claim 14(Currently Amended), Chauhan teaches A method, comprising: accessing a user history of previous actions performed by a user; using a machine learning model to learn user preferences for the user based on analyzing the user history or user profile information; creating a user-specific morph operator for the user that encodes the user preferences into a matrix (Chauhan, para 0088 discloses user specific features are learned from user profile using machine learning in a neural network “The associated machine learning model may be trained (either supervised or unsupervised) to create modified feature vectors for items that facilitate customized rankings of items for search results and/or listings through the search engine 106”; where para 0080-0083 further teaches profile or user historical data is being used for updating model “The user data 201 may include, but is not limited to, User IDs, profile data features 203, behavioral data features 205 (behavioral parameters), and/or other data features….”; para 0042 further discloses a default or generic vectors/embeddings/matrix transforming into personalized embedding by applying users’ identified preferences (morph operator)“a default vector may be preserved for use in general searches, administration, default rankings for items, and other uses, while periodically updated clusters of users and corresponding machine learning models may be used to generate cluster-specific vectors to rerank search results such that users in a particular cluster are provided ranked search results that are likely closer to the preferences of the users that belong in the particular cluster”);
a recommendation with one or more items to present that are related to the event and incorporate preferences of the user based on the personalized embedding(Chauhan, para 0093 discloses providing a search result based on modified/personalized embeddings which reflects user’s personal preferences “The search engine 106 may further provide a customized ranked list 315 and 317 based on the modified vectors 308 and 310, respectively. Each customized ranked list 315 and 317 may be presented to a user…. have a customized ranked list 315 presented in response to a search query that is customized based on the user data 201 and associated profile data and behavioral data” ).
Using the broadest reasonable interpretation consistent with the specification (paragraph 0030) as it would be interpreted by one of ordinary skill in the art, examiner is interpreting the limitation “user-specific morph operator” to mean identified user preference features which dictates user preferences and, the limitation “one or more items to present that are related to the event” to mean presenting items related to a query which is being triggered by an event.
But Chauhan does not explicitly disclose identifying an event performed by the user; using the user-specific morph operator to generate a personalized embedding for the event, wherein the personalized embedding for the event personalizes the event and items retrieved to the user preferences while keeping a corpus storing the items non personalized providing two-sided personalization on the event and item side while incurring a cost of one- sided personalization by using the matrix decomposition of the user-specific morph operator into multiple matrices; querying, using the personalized embedding for the event, a maximum inner product search (MIPS) structure that modifies retrievals for the event personalized to the user; and provide, in response to the querying,
However, in the same field of endeavor of content personalization LuVogt teaches identifying an event performed by the user; using the user-specific morph operator to generate a personalized embedding for the event(LuVogt, para 0125 discloses event (click event) is being identified and user vector or embeddings is getting updated with user preferences “If the user clicks on the content item D 502, the user vector 944 is combined with the representation of the content item D 502, the information regarding the click event, for example, the temporal metadata associated with the click event, and the user preferences in order to recalculate the user vector. Therefore, each time a user selects an item of content, the event generated by such selection affects the user model 3324 in real time”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of generating personalized embeddings for events of LuVogt into generic and personalized embeddings of Chauhan to produce an expected result of making embeddings for events. The modification would be obvious because one of ordinary skill in the art would be motivated to implement the process of updating user model/embeddings based on user selection (event) so that content recommendation to users would be in real-time on the basis of instantly updated user model (LuVogt, para 0011).
But Chauhan and LuVogt don’t explicitly teach wherein the personalized embedding for the event personalizes the event and items retrieved to the user preferences while keeping a corpus storing the items non personalized providing two-sided personalization on the event and item side while incurring a cost of one- sided personalization by using the matrix decomposition of the user-specific morph operator into multiple matrices; querying, using the personalized embedding for the event, a maximum inner product search (MIPS) structure that modifies retrievals for the event personalized to the user; and provide, in response to the querying,
However, in the same field of endeavor of query embeddings Thonet teaches wherein the personalized embedding for the event personalizes the event and items retrieved to the user preferences while keeping a corpus storing the items non personalized providing two-sided personalization on the event and item side while incurring a cost of one- sided personalization by using the matrix decomposition of the user-specific morph operator into multiple matrices(Thonet, claim 1 discloses providing two-sided personalization by updating the embedding first with user query and second with item “receiving training data including search instances and recommendation instances, each of the search instances including (a) a user profile, (b) a query comprising one or more query terms, and (c) an item………….. computing, based on the base embeddings and the respective embeddings obtained from each of the one or more convolutional layers: a first loss determined based on a prediction by the HGCN of an item based on a corresponding user profile and query; and a second loss determined based on a prediction by the HGCN of a query based on a corresponding user profile and item; and selectively updating at least one of the base embeddings of the HGCN based on the first loss and the second loss”; where items are being used for updating the base embedding for personalization but no update to items are being performed; where the limitation “by using the matrix decomposition of the user-specific morph operator into multiple matrices” is taught by prior art Kanno discussed later);
Using the broadest reasonable interpretation consistent with the specification (paragraph 0028 and 0033) as it would be interpreted by one of ordinary skill in the art, examiner is interpreting the limitation “providing two-sided personalization on the event and item side while incurring a cost of one- sided personalization” to mean
having both query and item embeddings in the same embedding space that represents user personalization embedding to achieve single cost of personalization by avoiding two separate embeddings and, the limitation “personalizes the event” to mean personalizing with the query information which is being triggered by an event.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of updating basic embedding for personalization of Thonet into the feature of generic and personal embeddings of Chauhan and LuVogt to produce an expected result of making embeddings personalized with query/intent features and its corresponding item features. The modification would be obvious because one of ordinary skill in the art would be motivated to determine the deficiencies (loss) in personalization in embeddings and updating the personalized embeddings accordingly (Thonet, abstract).
But Chauhan, LuVogt and Thonet don’t explicitly teach by using the matrix decomposition of the user-specific morph operator into multiple matrices; querying, using the personalized embedding for the event, a maximum inner product search (MIPS) structure that modifies retrievals for the event personalized to the user; and provide, in response to the querying,
However, in the same field of endeavor of query embeddings for personalization Kanno teaches by using the matrix decomposition of the user-specific morph operator into multiple matrices(Kanno, claim 6 (page 12 line 31-36) discloses decomposing user specific matrix (user-specific morph operator) into plurality of matrices for content personalization “A preference matrix in which user preference information for each item of each user is arranged in n rows and m columns, where n is the number of users and m is the number of items, is n rows and k columns (k is an integer smaller than m with 2 or more). A preference concept vector extraction unit that decomposes into a user's preference concept matrix and a preference concept matrix of items of k rows and m columns”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of decomposing user preference matrix of Kanno into the feature of item and query embeddings of Chauhan, LuVogt and Thonet to produce an expected result of embedding personalization. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the item recommendation by personalization(Kanno, claim 6).
But Chauhan, LuVogt, Thonet and Kanno don’t explicitly teach querying, using the personalized embedding for the event, a maximum inner product search (MIPS) structure that modifies retrievals for the event personalized to the user; and provide, in response to the querying,
However, in the same field of endeavor of query embeddings Shemesh teaches querying, using the personalized embedding for the event, a maximum inner product search (MIPS) structure that modifies retrievals for the event personalized to the user; and provide, in response to the querying (Shemesh, para 0065 discloses that querying using user vector/embeddings (personalized) and MIPS which performs search operations and obtain products which are similar by measures “Inner product search, or maximum inner product search, as used herein, refers to a search that, for a given user vector, performs a dot product of the user vector with all the candidate vectors, and selects a number of candidate vectors (e.g., top K out of all N candidate vectors) with maximum dot product. In this way, the inner product search relates to nearest neighbor search, which finds the point in a given set that is closest (or mostly similar) to a given point, only that, in this specific case, the closeness or similarity is defined in terms of an inner product similarity measure” ).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of using MIPS of Shemesh into the feature of generic and personal embeddings of Chauhan, LuVogt, Thonet and Kanno to produce an expected result of making embeddings personalized in real time. The modification would be obvious because one of ordinary skill in the art would be motivated to reduce the computational cost by filter-wise implementing MIPS(Shemesh, abstract).
Regarding claim 15 (Original), Chauhan, LuVogt, Thonet, Kanno and Shemesh teach all the limitations of claim 14 and Chauhan further teaches wherein the user-specific morph operator identifies the user preferences learned for the user (Chauhan, para 0088 discloses user specific features are learned from user profile using machine learning in a neural network “The associated machine learning model may be trained (either supervised or unsupervised) to create modified feature vectors for items that facilitate customized rankings of items for search results and/or listings through the search engine 106”; where para 0080-0083 further teaches profile or user historical data is being used for updating model “The user data 201 may include, but is not limited to, User IDs, profile data features 203, behavioral data features 205 (behavioral parameters), and/or other data features….”).
Regarding claim 17(Original), Chauhan, LuVogt, Thonet, Kanno and Shemesh teach all the limitations of claim 14 and Chauhan further teaches wherein the user-specific morph operator is a function based on the user history of previous actions performed by the user or user profile information (Chauhan, para 0080-0083 teaches profile or user historical data is being used for updating model or user specific information which is being used for personalization are from user profile and behavioral data “The user data 201 may include, but is not limited to, User IDs, profile data features 203, behavioral data features 205 (behavioral parameters), and/or other data features….”).
Claim 16 and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Chauhan, Sejal et al (PGPUB Document No. 20220067052), hereafter referred as to “Chauhan”, in view of LuVogt, Chris et al (PGPUB Document No. 20130290110), hereafter, referred to as “LuVogt”, in view of Thonet, Thibaut et al (US Patent No. 12399945), hereafter, referred to as “Thonet”, in view of Kanno, Kyota (WIPO Publication WO2011049037), hereafter, referred to as “Kanno”, in view of Shemesh, Ariel et al (PGPUB Document No. 20200005119), hereafter, referred to as “Shemesh”, in further view of Hsieh, Cheng-Kang et al (PGPUB Document No. 20210174164), hereafter, referred to as “Hsieh”.
Regarding claim 16(Original), Chauhan, LuVogt, Thonet, Kanno and Shemesh teach all the limitations of claim 14 but don’t explicitly teach wherein the user-specific morph operator is a separate machine learning model or a matrix.
However, in the same field of endeavor of query embeddings Hsieh teaches wherein the user-specific morph operator is a separate machine learning model or a matrix (Hsieh, Fig. 1 and para 0064-0065 disclose user personal feature extraction using metric learning method “Through collaborative metric learning, the system and method may be able to establish an understanding of user preferences and affinities to content and other users by using an existing user profile as a look-alike reference”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of having machine learning model for user preference features of Hsieh into the feature of generic and personal embeddings of Chauhan, LuVogt, Thonet, Kanno and Shemesh to produce an expected result of making embeddings personalized in real time. The modification would be obvious because one of ordinary skill in the art would be motivated to use content-centric predictive personalization to recommend contents to new users in real or run time where no prior data for user personalization is available(Hsieh, para 0030).
Regarding claim 18(Original), Chauhan, LuVogt, Thonet, Kanno and Shemesh teach all the limitations of claim 14 but don’t explicitly teach further comprising: creating an intermediate user embedding of the user history or the user profile information, wherein the intermediate user embedding is an aggregate of the previous actions performed by the user.
However, in the same field of endeavor of query embeddings Hsieh teaches further comprising: creating an intermediate user embedding of the user history or the user profile information(Hsieh, element S132 of Fig. 12 and para 0121 disclose user profile information is being used for creating user embedding “processing a user data comprised of user feature data as input to a user neural network model and yielding a user embedding S132”), wherein the intermediate user embedding is an aggregate of the previous actions performed by the user (Hsieh, para 0164 discloses embedding or personalization can be done using users’ historical activities (user feature data which is being inputted for embedding) over a time period or aggregation of actions “The method can be used for deep personalization utilizing a history of previously observed user activity. … personalization may be based on a fixed time window such as activity in the last 10 minutes, hour, day, or week”; here the examiner interprets “intermediate user embedding” as any embedding which is to be further processed or updated).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of creating an intermediate embedding with user preference features of Hsieh into the feature of generic and personal embeddings of Chauhan, LuVogt, Thonet, Kanno and Shemesh to produce an expected result of making embeddings personalized in real time. The modification would be obvious because one of ordinary skill in the art would be motivated to use content-centric predictive personalization to recommend contents to new users in real or run time where no prior data for user personalization is available(Hsieh, para 0030).
Regarding claim 19(Original), Chauhan, LuVogt, Thonet, Kanno, Shemesh and Hsieh all the limitations of claim 18 and Hsieh further teaches wherein the intermediate user embedding is used to create the user-specific morph operator (Hsieh, Fig. 1 and para 0064-0065 further disclose creating user specific morph operator (a metric space to encode user content preferences) is being created using intermediate/user embedding (element 130) “The matchmaking neural network 140 (MmNN) of a preferred embodiment functions as a computational model that enhances a shared mapping of content embeddings from the C-NN 120 and user embeddings from the U-NN 130. The MmNN 140 is preferably a system that learns a joint metric space to encode user content preferences through item proximity”).
Regarding claim 20(Original), Chauhan, LuVogt, Thonet, Kanno, Shemesh and Hsieh all the limitations of claim 18 and Hsieh further teaches wherein the intermediate user embedding is generated offline using a different machine learning model and stored for the user(Hsieh, para 0122 discloses embedding can be generated ahead of time and stored “The user data may be processed ahead of time or on demand in response to a triggering event (such as the receipt of a query input associated with a user). In some implementations, user shared-item embeddings are calculated and stored as part of the matchmaking data model”; where element S132 of Fig. 12 and para 0121 disclose embedding is being generated using neural network or machine learning),
and wherein the intermediate user embedding is used to create the user-specific morph operator in real time (Hsieh, Fig. 1 and para 0064-0065 further disclose creating user specific morph operator (a metric space to encode user content preferences) is being created using intermediate/ user embedding (element 130) “The matchmaking neural network 140 (MmNN) of a preferred embodiment functions as a computational model that enhances a shared mapping of content embeddings from the C-NN 120 and user embeddings from the U-NN 130. The MmNN 140 is preferably a system that learns a joint metric space to encode user content preferences through item proximity”; para 0122 further teaches the intermediate embedding which creates morph operator (a metric space to encode user content preferences) process the user data and process of user data takes place by triggering event such as upon receipt of query “The user data may be processed ahead of time or on demand in response to a triggering event (such as the receipt of a query input associated with a user). In some implementations, user shared-item embeddings are calculated and stored as part of the matchmaking data model”).
Response to Arguments
I. 35 U.S.C §103
Applicant’s arguments filed on 12/5/2025 have been fully considered but are
moot because the independent claim 1 and 14 have been amended with newly added
features which applicant’s arguments are directed towards. Since claims have been amended with new features, a new ground of rejection is presented.
Conclusion
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ABDULLAH A DAUD whose telephone number is (469)295-9283. The examiner can normally be reached M~F: 9:30 am~6:30 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, Amy Ng can be reached at 571-270-1698. 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.
/ABDULLAH A DAUD/Examiner, Art Unit 2164
/MARK E HERSHLEY/Primary Examiner, Art Unit 2164