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 .
Information Disclosure
IDS Submitted on 02/24/2026 and 11/05/2025 have been considered by the examiner.
Response to Amendment
This Office action is in response to Applicant's amendment filed on 2/2/2026.
Claim 1-20 are pending. Claim 1, 15 and 20 are amended. Claim 1-20 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, 4-7, 15 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Moon, Changsung et al(PGPUB Document No. 20230170092), hereafter referred as to “Moon”, in view of Wang, Zhenrui et al (PGPUB Document No. 20220245162), hereafter, referred to as “Wang”, in further view of Sheshansh, Satyam et al (PGPUB Document No. 20230368130), hereafter, referred to as “Sheshansh”.
Regarding Claim 1 (Currently Amended), Moon teaches A system comprising one or more processors and at least one memory storing processor-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising(Moon, para 0014 teaches a system with processor, memory and data storages): receiving a search query; in response to the search query, receiving a plurality of search results associated with a plurality of relevance scores in an initial ranking order(Moon, para 0052 plurality of similarity scores are getting determined to rank each output (medical indication) in response to query “….cosine similarity score(s) may be generated between the examination order vector and the aggregated medical indication vector and/or the individual medical indication vector and used to rank the various medical indications in terms based on their probability of being relevant or applicable responsible to the input feature set 1 and input feature set 2 information”; here the examiner interprets ranking for each output as initial ranking of data object/output (medical indication));
generating, based at least in part on the relevance score subset, a per-measure optimized ranking order of the plurality of search results according to the relevance measure(Moon, para 0055 discloses relevance scores are being compared against a threshold for ranking of output (medical indication) based on each determined score “The similarity scores (e.g., cosine similarity scores) are indicative of the probabilities that respective ones of the medical indications are applicable to an examination order. Thus, in some embodiments, when a particular medical indication has a similarity score corresponding to a probability that exceeds a defined threshold for an examination order”);
But Moon does not explicitly teach re-ranking the plurality of search results by: generating a plurality of ranking comparison scores by: determining, from the plurality of relevance scores, a relevance score subset that is associated with a relevance measure; and generating the plurality of ranking comparison scores representing ranking differences between the initial ranking order and the per-measure optimized ranking order associated with the relevance measure; and generating, by inputting the plurality of ranking comparison scores into a multi-measure ranking optimization machine learning model, a multi-measure optimized ranking order associated with the plurality of search results, wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs; and outputting the plurality of search results according to the multi-measure optimized ranking order to a client computing entity.
However, in the same field of endeavor of content ranking by similarity Wang teaches re-ranking the plurality of search results by(Wang, para 0071 discloses re-ranking the initial ranked list “Item ranking computing device 102 may then re-rank the initial set of recommended items based on the final ranking scores. Item ranking computing device 102 may transmit ranked search results 312 to web server 104, where ranked search results 312 identifies the re-ranked set of recommended items”): generating a plurality of ranking comparison scores by: determining, from the plurality of relevance scores, a relevance score subset that is associated with a relevance measure; and generating the plurality of ranking comparison scores representing ranking differences between the initial ranking order and the per-measure optimized ranking order associated with the relevance measure(Wang, Fig. 4 and para 0078 disclose a single ranking score is being generated based on plurality of individual ranking scores “Blend model engine 408 receives relevance scores 403 from relevance model engine 402 and SR scores 405 from SRS model engine 404, and determines blended relevance scores 409 based on relevance scores 403 and SR scores 405”); and generating, by inputting the plurality of ranking comparison scores into a multi-measure ranking optimization machine learning model, a multi-measure optimized ranking order associated with the plurality of search results(Wang, Fig. 4 and para 0078 further teach multiple measure raking order (scores) such as relevance score and semantic relevance score are getting inputted into another learning model (element 408 of Fig. 4A) for an optimized or final ranking scoring “blend model engine 408 may determine a blended relevance score 409 for an item based on averaging the relevance score 403 and SR score 405 for the item. In some examples, blend model engine 408 applies a weight to each of the relevance score 403 and SR score 405, and combines the weighted scores to generate blended relevance score 40”); and outputting the plurality of search results according to the multi-measure optimized ranking order to a client computing entity(Wang, para 0080 teaches outputting a ranked item result based on combined individual optimized final raking score “Final score determination engine 412 receives blended relevance score 409 from blend model engine 408, and engagement score 407 from engagement model engine 406. ……… Final score determination engine 412 may determine a final ranking score for each of the recommended items 399, and generate ranked search results 312 based on the final ranking scores”).
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 combining plurality of ranking scores to form a single ranking scores of Wang into the feature of generating plurality of relevance scores of Moon to produce an expected result of generating an optimized ranking. The modification would be obvious because one of ordinary skill in the art would be motivated to rank more relevant item ahead of less relevant items by using trained machine learning model which generates features for relevance determination(Wang, para 0003).
But Moon and Wang don’t explicitly teach wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs;
However, in the same field of endeavor of content ranking using pairwise comparison Sheshansh teaches wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs(Sheshansh, para 0035 discloses machine learning model to optimize ranking by using pairwise ranking algorithm where, probabilistic function optimizes the number of swaps of contents “the ranking-based ML algorithm of order prioritization engine 116 includes a pairwise ranking algorithm. In brief, the pairwise ranking algorithm compares two orders and ranks the two orders based on their priority. The pairwise ranking approaches employ a probabilistic cost function which considers a pair of orders in a dataset at a time in the loss function to learn how to rank the orders (e.g., to estimate which of the two orders is more relevant than the other). The goal of the loss function is to minimize the number of swaps (e.g., the number of inversions) required to correct an incorrect ordering of candidate orders”);
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 optimizing ranked content based on their priority using pairwise ranking algorithm the of Sheshansh into the feature of generating plurality of relevance scores and combining of ranking scores of Moon and Wang to produce an expected result of optimizing contents differently ordered. The modification would be obvious because one of ordinary skill in the art would be motivated to rank more relevant content item by minimizing the number of swaps required to correct an incorrect ordering of candidate orders(Sheshansh, para 0035).
Regarding Claim 4(Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 2 and Moon further discloses the operation further comprise: generating a plurality of query feature vectors based at least in part on the search query, wherein the plurality of query feature vectors comprises one or more of query embedding vectors and query-item relevance vectors(Moon, para 0052 discloses query embedding is being performed based on user’s search input (free-text query vector) and further teaches query item relevance scores are getting considered “The AI medical indication prediction engine 230 may generate an output comprising one or more ranked medical indications based on similarity scores. The similarity scores may be indicative of a similarity between the examination order vector, which comprises an aggregation of the input variable vectors and the free-text query vector”).
Regarding Claim 5(Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 4 and Moon further discloses the operation further comprise: generating the initial ranking order based at least in part on the plurality of user feature vectors and the plurality of query feature vectors(Moon, para 0052 discloses relevance scores which indicates vectors such as query vector and medical indication vector are being used for ranking “The AI medical indication prediction engine 230 may generate an output comprising one or more ranked medical indications based on similarity scores. The similarity scores may be indicative of a similarity between the examination order vector, which comprises an aggregation of the input variable vectors and the free-text query vector, and an aggregated medical indication vector …” ; where para 0046 discloses input features to machine leaning model for prediction is having user information embeddings ).
Regarding Claim 6(Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 1 and Wang further teaches wherein the plurality of relevance scores comprises a plurality of textual relevance scores (Wang, Fig. 4A and para 0078 discloses plurality of relevance scores such as “relevance score” (element 403 of Fig. 4A) and syntactical relevance score (element 405 of Fig. 4A) ), a plurality of engagement relevance scores, and a plurality of outcome relevance scores(Wang, Fig. 4A and para 0074 disclose engagement relevance or outcome relevance score engine 406 may obtain search request 397, recommended items 399, and user session data 320 (e.g., for the user generating search request 310) from database 116, and may apply a trained engagement model to the search request 397, recommended items 399, and user session data 320 to generate a engagement score 407 for each item”).
Regarding Claim 7 (Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 4 and Moon further discloses the operation further comprise: generating a plurality of query feature vectors based at least in part on the search query data object(Moon, para 0052 discloses query vector and medical indication vector are being used for ranking “The AI medical indication prediction engine 230 may generate an output comprising one or more ranked medical indications based on similarity scores. The similarity scores may be indicative of a similarity between the examination order vector, which comprises an aggregation of the input variable vectors and the free-text query vector, and an aggregated medical indication vector …” );
determining search result metadata that are associated with the plurality of search results(Moon, para 0046 further discloses metadata such as information about patient is being used “Input feature set 1 205 may represent patient, provider, and/or order information. The patient information may include, but is not limited to, age, gender, problem list, encounter diagnosis, patient class, and/or a medical center department”); and generating the plurality of textual relevance scores based at least in part on search result metadata and the plurality of query feature vectors(Moon, para 0052 discloses relevance scores which indicates vectors such as query vector and medical indication vector are being used for ranking “The similarity scores may be indicative of a similarity between the examination order vector, which comprises an aggregation of the input variable vectors and the free-text query vector, and an aggregated medical indication vector …” ; where para 0046 discloses input features to machine leaning model for prediction is having user information embeddings (query feature vector) “Input feature set 1 205 may represent patient, provider, and/or order information. The patient information may include, but is not limited to, age, gender, problem list, encounter diagnosis, patient class, and/or a medical center department”).
Regarding Claim 15 (Currently Amended), Moon teaches A computer-implemented method comprising: receiving, one or more processors; in response to the search query, receiving, using the one or more processors; a plurality of search results associated with a plurality of relevance scores in an initial ranking order(Moon, para 0052 plurality of similarity scores are getting determined to rank each output (medical indication) in response to query “….cosine similarity score(s) may be generated between the examination order vector and the aggregated medical indication vector and/or the individual medical indication vector and used to rank the various medical indications in terms based on their probability of being relevant or applicable responsible to the input feature set 1 and input feature set 2 information”; here the examiner interprets ranking for each output as initial ranking of data object/output (medical indication));
and generating the plurality of ranking comparison scores representing ranking differences between the initial ranking order and the per-measure optimized ranking order associated with the relevance measure(Moon, para 0055 discloses relevance scores are being compared against a threshold for ranking of output (medical indication) based on each determined score “The similarity scores (e.g., cosine similarity scores) are indicative of the probabilities that respective ones of the medical indications are applicable to an examination order. Thus, in some embodiments, when a particular medical indication has a similarity score corresponding to a probability that exceeds a defined threshold for an examination order”);
But Moon does not explicitly teach re-ranking the plurality of search results by: generating, by the one or more processors, a plurality of ranking comparison scores by: determining, from the plurality of relevance scores, a relevance score subset that is associated with a relevance measure;
generating, based at least in part on the relevance score subset, a per-measure optimized ranking order of search results according to the relevance measure; and generating, by inputting the plurality of ranking comparison scores into a multi-measure ranking optimization machine learning model, a multi-measure optimized ranking order associated with the plurality of search results; wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs; and outputting the plurality of search results according to the multi-measure optimized ranking order to a client computing entity.
However, in the same field of endeavor of content ranking by similarity Wang teaches re-ranking the plurality of search results by(Wang, para 0071 discloses re-ranking the initial ranked list “Item ranking computing device 102 may then re-rank the initial set of recommended items based on the final ranking scores. Item ranking computing device 102 may transmit ranked search results 312 to web server 104, where ranked search results 312 identifies the re-ranked set of recommended items”): generating, by the one or more processors, a plurality of ranking comparison scores by: determining, from the plurality of relevance scores, a relevance score subset that is associated with a relevance measure; generating, based at least in part on the relevance score subset, a per-measure optimized ranking order of search results according to the relevance measure(Wang, Fig. 4 and para 0078 disclose a single ranking score is being generated based on plurality of individual ranking scores “Blend model engine 408 receives relevance scores 403 from relevance model engine 402 and SR scores 405 from SRS model engine 404, and determines blended relevance scores 409 based on relevance scores 403 and SR scores 405”); and generating, by inputting the plurality of ranking comparison scores into a multi-measure ranking optimization machine learning model, a multi-measure optimized ranking order associated with the plurality of search results(Wang, Fig. 4 and para 0078 further teach multiple measure raking order (scores) such as relevance score and semantic relevance score are getting inputted into another learning model (element 408 of Fig. 4A) for an optimized or final ranking scoring “blend model engine 408 may determine a blended relevance score 409 for an item based on averaging the relevance score 403 and SR score 405 for the item. In some examples, blend model engine 408 applies a weight to each of the relevance score 403 and SR score 405, and combines the weighted scores to generate blended relevance score 40”); and outputting the plurality of search results according to the multi-measure optimized ranking order to a client computing entity(Wang, para 0080 teaches outputting a ranked item result based on combined individual optimized final raking score “Final score determination engine 412 receives blended relevance score 409 from blend model engine 408, and engagement score 407 from engagement model engine 406. ……… Final score determination engine 412 may determine a final ranking score for each of the recommended items 399, and generate ranked search results 312 based on the final ranking scores”).
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 combining plurality of ranking scores to form a single ranking scores into the feature of generating plurality of relevance scores of Moon to produce an expected result of generating an optimized ranking. The modification would be obvious because one of ordinary skill in the art would be motivated to rank more relevant item ahead of less relevant items by using trained machine learning model which generates features for relevance determination(Wang, para 0003).
But Moon and Wang don’t explicitly teach wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs;
However, in the same field of endeavor of content ranking using pairwise comparison Sheshansh teaches wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs(Sheshansh, para 0035 discloses machine learning model to optimize ranking by using pairwise ranking algorithm where, probabilistic function optimizes the number of swaps of contents “the ranking-based ML algorithm of order prioritization engine 116 includes a pairwise ranking algorithm. In brief, the pairwise ranking algorithm compares two orders and ranks the two orders based on their priority. The pairwise ranking approaches employ a probabilistic cost function which considers a pair of orders in a dataset at a time in the loss function to learn how to rank the orders (e.g., to estimate which of the two orders is more relevant than the other). The goal of the loss function is to minimize the number of swaps (e.g., the number of inversions) required to correct an incorrect ordering of candidate orders”);
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 optimizing ranked content based on their priority using pairwise ranking algorithm the of Sheshansh into the feature of generating plurality of relevance scores and combining of ranking scores of Moon and Wang to produce an expected result of optimizing contents differently ordered. The modification would be obvious because one of ordinary skill in the art would be motivated to rank more relevant content item by minimizing the number of swaps required to correct an incorrect ordering of candidate orders(Sheshansh, para 0035).
Regarding Claim 20 (Currently Amended), Moon teaches One or more non-transitory computer-readable storage media storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising(Moon, para 0014 teaches a system with processor, memory and data storages): receiving a search query; in response to the search query, receiving a plurality of search results associated with a plurality of relevance scores in an initial ranking order(Moon, para 0052 plurality of similarity scores are getting determined to rank each output (medical indication) in response to query “….cosine similarity score(s) may be generated between the examination order vector and the aggregated medical indication vector and/or the individual medical indication vector and used to rank the various medical indications in terms based on their probability of being relevant or applicable responsible to the input feature set 1 and input feature set 2 information”; here the examiner interprets ranking for each output as initial ranking of data object/output (medical indication));
generating, based at least in part on the relevance score, a per-measure optimized ranking order of the plurality of search results according to the relevance measure(Moon, para 0055 discloses relevance scores are being compared against a threshold for ranking of output (medical indication) based on each determined score “The similarity scores (e.g., cosine similarity scores) are indicative of the probabilities that respective ones of the medical indications are applicable to an examination order. Thus, in some embodiments, when a particular medical indication has a similarity score corresponding to a probability that exceeds a defined threshold for an examination order”);
But Moon does not explicitly teach re-ranking the plurality of search results by: generating a plurality of ranking comparison scores by: determining, from the plurality of relevance
and generating, by inputting the plurality of ranking comparison scores into a multi-measure ranking optimization machine learning model, a multi-measure optimized ranking order associated with the plurality of search results; wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs; and outputting the plurality of search results according to the multi-measure optimized ranking order to a client computing entity.
However, in the same field of endeavor of content ranking by similarity Wang teaches re-ranking the plurality of search results by(Wang, para 0071 discloses re-ranking the initial ranked list “Item ranking computing device 102 may then re-rank the initial set of recommended items based on the final ranking scores. Item ranking computing device 102 may transmit ranked search results 312 to web server 104, where ranked search results 312 identifies the re-ranked set of recommended items”): generating a plurality of ranking comparison scores by: determining, from the plurality of relevance(Wang, Fig. 4 and para 0078 disclose a single ranking score is being generated based on plurality of individual ranking scores “Blend model engine 408 receives relevance scores 403 from relevance model engine 402 and SR scores 405 from SRS model engine 404, and determines blended relevance scores 409 based on relevance scores 403 and SR scores 405”);
and generating, by inputting the plurality of ranking comparison scores into a multi-measure ranking optimization machine learning model, a multi-measure optimized ranking order associated with the plurality of search results(Wang, Fig. 4 and para 0078 further teach multiple measure raking order (scores) such as relevance score and semantic relevance score are getting inputted into another learning model (element 408 of Fig. 4A) for an optimized or final ranking scoring “blend model engine 408 may determine a blended relevance score 409 for an item based on averaging the relevance score 403 and SR score 405 for the item. In some examples, blend model engine 408 applies a weight to each of the relevance score 403 and SR score 405, and combines the weighted scores to generate blended relevance score 40”); and outputting the plurality of search results according to the multi-measure optimized ranking order to a client computing entity(Wang, para 0080 teaches outputting a ranked item result based on combined individual optimized final raking score “Final score determination engine 412 receives blended relevance score 409 from blend model engine 408, and engagement score 407 from engagement model engine 406. ……… Final score determination engine 412 may determine a final ranking score for each of the recommended items 399, and generate ranked search results 312 based on the final ranking scores”).
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 combining plurality of ranking scores to form a single ranking scores into the feature of generating plurality of relevance scores of Moon to produce an expected result of generating an optimized ranking. The modification would be obvious because one of ordinary skill in the art would be motivated to rank more relevant item ahead of less relevant items by using trained machine learning model which generates features for relevance determination(Wang, para 0003).
But Moon and Wang don’t explicitly teach wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs;
However, in the same field of endeavor of content ranking using pairwise comparison Sheshansh teaches wherein the multi-measure ranking optimization machine learning model computes a plurality of pairwise relevance probabilities between a plurality of pairs of the plurality of search results, and optimizes the relevance measure based on a rank swap between one or more of the plurality of pairs (Sheshansh, para 0035 discloses machine learning model to optimize ranking by using pairwise ranking algorithm where, probabilistic function optimizes the number of swaps of contents “the ranking-based ML algorithm of order prioritization engine 116 includes a pairwise ranking algorithm. In brief, the pairwise ranking algorithm compares two orders and ranks the two orders based on their priority. The pairwise ranking approaches employ a probabilistic cost function which considers a pair of orders in a dataset at a time in the loss function to learn how to rank the orders (e.g., to estimate which of the two orders is more relevant than the other). The goal of the loss function is to minimize the number of swaps (e.g., the number of inversions) required to correct an incorrect ordering of candidate orders”);
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 optimizing ranked content based on their priority using pairwise ranking algorithm the of Sheshansh into the feature of generating plurality of relevance scores and combining of ranking scores of Moon and Wang to produce an expected result of optimizing contents differently ordered. The modification would be obvious because one of ordinary skill in the art would be motivated to rank more relevant content item by minimizing the number of swaps required to correct an incorrect ordering of candidate orders(Sheshansh, para 0035).
Claim 2-3, 8 and 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Moon, Changsung et al(PGPUB Document No. 20230170092), hereafter referred as to “Moon”, in view of Wang, Zhenrui et al (PGPUB Document No. 20220245162), hereafter, referred to as “Wang”, in view of Sheshansh, Satyam et al (PGPUB Document No. 20230368130), hereafter, referred to as “Sheshansh”, in further view of Gao, Tianshi et al (US Patent No. 10943178), hereafter, referred to as “Gao”.
Regarding Claim 2 (Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 1 but don’t explicitly teach the operation further comprise: receiving a user profile associated with the search query, wherein the user profile data object comprises user profile metadata; and generating a plurality of user feature vectors associated with the user profile based at least in part on the user profile metadata.
However, in the same field of endeavor of content ranking by similarity Gao teaches the operation further comprise: receiving a user profile associated with the search query, wherein the user profile data object comprises user profile metadata; and generating a plurality of user feature vectors associated with the user profile based at least in part on the user profile metadata(Gao, col 13: 4-8 discloses vector or embedding of users’ profile data such as user’s demographic feature “Based on the obtained information identifying interactions by users with content, the online system 140 generates 310 an embedding of each user and maintains the embeddings in association with their corresponding users. The embedding of to a user has multiple dimensions, with different dimensions corresponding to information about the user….. the online system 140 includes an embedding corresponding to a user in a user profile maintained by the online system 140 for the 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 personalized prediction-based content ranking of Gao into content ranking of Moon, Wang and Sheshansh to produce an expected result of generating personalized ranked contents by prediction. The modification would be obvious because one of ordinary skill in the art would be motivated to correct the limitation (bias only towards interaction only) of model training with user interaction data by using additional similarity measure between user embedding with content embedding for selection of items(Gao, col 14: 59-65 & col 15:10-15).
Regarding Claim 3(Previously Presented), Moon, Wang, Sheshansh and Gao teach all the limitations of claim 2 and Gao further teaches wherein the plurality of user feature vectors comprises one or more of user socio-economics embedding vectors, user demographics characteristics vectors, user search history embedding vectors, and user medical history embedding vectors(Gao, col 13: 11:14 discloses vector or embedding of users’ profile data such as user’s demographic feature “while other dimensions of the embedding describe characteristics of the user….….the online system 140 includes an embedding corresponding to a user in a user profile maintained by the online system 140 for the user”; where col 6:25-27 discloses users profile having demographic information).
Regarding Claim 8 (Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 4 and Moon further discloses receiving search result selection metadata associated with the plurality of search results(Moon, para 0034 discloses retrieving plurality of search result data (possible medical indication) for selection “information associated with a patient, such as age, gender, and an order description, when presenting a health care services provider (“provider”) with a list of possible medical indications for an order. This can result in the provider needing to review numerous possible medical indications to select one that best supports a particular order”);
But Moon, Wang and Sheshansh don’t explicitly teach generating one or more attractiveness variables, one or more examination variables, and one or more satisfaction variables associated with the plurality of search results based at least in part on the search result selection metadata; and generating the plurality of engagement relevance scores based at least in part on inputting the one or more attractiveness variables, the one or more examination variables, and the one or more satisfaction variables to an engagement relevance machine learning model.
However, in the same field of endeavor of content ranking by similarity Gao teaches generating one or more attractiveness variables(Gao, col 10:32-35 discloses attractiveness variable data or how likely a user associated with the profile would interact with the output content “the content selection module 230 includes one or more models that determine measures of relevance of various content items to the user based on characteristics of the user by the online system 140 and based on the user's affinity for different content items”), one or more examination variables(Moon, para 0034 discloses “examination variable data objects” such as users’ clinical diagnostic test information in the query “information associated with a patient, such as age, gender, and an order description, when presenting a health care services provider (“provider”) with a list of possible medical indications”), and one or more satisfaction variables associated with the plurality of search results based at least in part on the search result selection metadata(Gao, col 10:34-35 discloses satisfaction variable data or a measure of users’ affinity for output content items “determine measures of relevance of various content items to the user based on characteristics of the user by the online system 140 and based on the user's affinity for different content items”);
and generating the plurality of engagement relevance scores based at least in part on inputting the one or more attractiveness variables(Gao, col 10:32-35 discloses attractiveness variable data or how likely a user associated with the profile would interact with the output content “the content selection module 230 includes one or more models that determine measures of relevance of various content items to the user based on characteristics of the user by the online system 140 and based on the user's affinity for different content items”), the one or more examination variables, and the one or more satisfaction variables (Gao, col 10:34-35 discloses satisfaction variable data or a measure of users’ affinity for output content items “determine measures of relevance of various content items to the user based on characteristics of the user by the online system 140 and based on the user's affinity for different content items”) to an engagement relevance machine learning model(Gao, col 4:31-36 discloses “engagement relevance score” a score which indicates user likeliness to engage with the output content using similarity scores which indicates users’ likeliness (engagement) to interact with the provided contents “the online system includes the applies one or more selection processes to content items that ranks content items based on their corresponding combinations of likelihoods of the viewing user performing interactions identified by objectives included in various content items and measures of similarity of the embedding of the viewing user to embeddings for interactions identified by objectives included in various content items”; where Wang in Fig. 4A and para 0078 disclose use of machine learning model for generating engagement relevance score).
Using the broadest reasonable interpretation consistent with the specification (paragraph 0093-0094 and 0099-0100) as it would be interpreted by one of ordinary skill in the art, examiner is interpreting the limitation “attractiveness variables” to mean at least a score which measures how likely a user associated with the profile would interact with the output content, “satisfaction variables” to mean at least a measure of user's affinity for content items.
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 personalized prediction-based content ranking of Gao into content ranking of Moon, Wang and Sheshansh to produce an expected result of generating personalized ranked contents by prediction. The modification would be obvious because one of ordinary skill in the art would be motivated to correct the limitation (bias only towards interaction only) of model training with user interaction data by using additional similarity measure between user embedding with content embedding for selection of items(Gao, col 14: 59-65 & col 15:10-15).
Regarding Claim 16 (Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 15 but don’t explicitly teach further comprising: receiving a user profile associated with the search quer
However, in the same field of endeavor of content ranking by similarity Gao teaches further comprising: receiving a user profile associated with the search querplurality of user feature vectors associated with the user profile based at least in part on the user profile metadata (Gao, col 13: 4-8 discloses vector or embedding of users’ profile data such as user’s demographic feature “Based on the obtained information identifying interactions by users with content, the online system 140 generates 310 an embedding of each user and maintains the embeddings in association with their corresponding users. The embedding of to a user has multiple dimensions, with different dimensions corresponding to information about the user….. the online system 140 includes an embedding corresponding to a user in a user profile maintained by the online system 140 for the 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 personalized prediction-based content ranking of Gao into content ranking of Moon, Wang and Sheshansh to produce an expected result of generating personalized ranked contents by prediction. The modification would be obvious because one of ordinary skill in the art would be motivated to correct the limitation (bias only towards interaction only) of model training with user interaction data by using additional similarity measure between user embedding with content embedding for selection of items(Gao, col 14: 59-65 & col 15:10-15).
Regarding Claim 17(Previously Presented), Moon, Wang, Sheshansh and Gao teach all the limitations of claim 16 and Gao further teaches wherein the plurality of user feature vectors comprises one or more of user socio-economics embedding vectors, user demographics characteristics vectors, user search history embedding vectors, and user medical history embedding vectors (Gao, col 13: 11:14 discloses vector or embedding of users’ profile data such as user’s demographic feature “while other dimensions of the embedding describe characteristics of the user….….the online system 140 includes an embedding corresponding to a user in a user profile maintained by the online system 140 for the user”; where col 6:25-27 discloses users profile having demographic information).
Regarding Claim 18(Previously Presented), Moon, Wang, Sheshansh and Gao teach all the limitations of claim 16 and Moon further discloses further comprising: generating a plurality of query feature vectors based at least in part on the search query, wherein the plurality of query feature vectors comprises one or more of query embedding vectors and query-item relevance vectors (Moon, para 0052 discloses query embedding is being performed based on user’s search input (free-text query vector) and further teaches query item relevance scores are getting considered “The AI medical indication prediction engine 230 may generate an output comprising one or more ranked medical indications based on similarity scores. The similarity scores may be indicative of a similarity between the examination order vector, which comprises an aggregation of the input variable vectors and the free-text query vector”).
Regarding Claim 19(Previously Presented), Moon, Wang, Sheshansh and Gao teach all the limitations of claim 18 and Moon further teaches further comprising: generating the initial ranking order based at least in part on the plurality of user feature vectors and the plurality of query feature vectors (Moon, para 0052 discloses relevance scores which indicates vectors such as query vector and medical indication vector are being used for ranking “The AI medical indication prediction engine 230 may generate an output comprising one or more ranked medical indications based on similarity scores. The similarity scores may be indicative of a similarity between the examination order vector, which comprises an aggregation of the input variable vectors and the free-text query vector, and an aggregated medical indication vector …” ; where para 0046 discloses input features to machine leaning model for prediction is having user information embeddings ).
Claim 9-11 are rejected under 35 U.S.C. 103 as being unpatentable over Moon, Changsung et al(PGPUB Document No. 20230170092), hereafter referred as to “Moon”, in view of Wang, Zhenrui et al (PGPUB Document No. 20220245162), hereafter, referred to as “Wang”, in view of Sheshansh, Satyam et al (PGPUB Document No. 20230368130), hereafter, referred to as “Sheshansh”, in further view of Bhatia, Vidit et al (PGPUB Document No. 20200104395), hereafter, referred to as “Bhatia”.
Regarding Claim 9 (Previously Presented), Moon, Wang and Sheshansh teach all the limitations of claim 6 but don’t explicitly teach wherein the plurality of engagement relevance scores comprises a plurality of immediate engagement relevance
However in the same field of endeavor of content interaction data capturing Bhatia teaches wherein the plurality of engagement relevance scores comprises a plurality of immediate engagement relevance (Bhatia, para 0111 discloses the time taken to react or interact (immediate/delayed engagement) for users upon receiving a content is being measured “the user device 102/103 and/or meet-up interface 303 may include the timer that starts upon receiving the meet-up recommendation 108. In some embodiments, the predetermined period of time may include any suitable time period for user interaction with the meet-up interface 303”; where Wang in para 0078 discloses engagement score).
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 time taken for users to interact of Bhatia into content ranking of Moon, Wang and Sheshansh to produce an expected result of presenting contents to users based on the likeliness of contents to be interacted. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the accuracy of user segment by using uniform users embeddings to determine holistic similarities of users in a group(Bhatia, para 0022).
Regarding Claim 10 (Previously Presented), Moon, Wang, Sheshansh and Bhatia teach all the limitations of claim 9 and Bhatia further teaches receiving search result completion metadata associated with the plurality of search results; and generating the plurality of immediate engagement relevance scores based at least in part on the search result completion metadata(Bhatia, para 0111 discloses the time taken to react or interact (immediate/delayed engagement) for users upon receiving a content is being measured by metadata (take taken to interact) “the user device 102/103 and/or meet-up interface 303 may include the timer that starts upon receiving the meet-up recommendation 108. In some embodiments, the predetermined period of time may include any suitable time period for user interaction with the meet-up interface 303”).
Regarding Claim 11(Previously Presented), Moon, Wang, Sheshansh and Bhatia teach all the limitations of claim 9 and Bhatia further teaches wherein the operations further comprise: determining a post-search observation time period that is associated with the plurality of search results (Bhatia, para 0111 discloses the time taken to react or interact for users upon receiving a content is being measured “the user device 102/103 and/or meet-up interface 303 may include the timer that starts upon receiving the meet-up recommendation 108. In some embodiments, the predetermined period of time may include any suitable time period for user interaction with the meet-up interface 303”); receiving a user profile (Moon, para 0034 discloses retrieving users’ clinical diagnostic test information with user’ profile data such as age, gender etc. “information associated with a patient, such as age, gender, and an order description, when presenting a health care services provider (“provider”) with a list of possible medical indications”)
Bhatia teaches and the post-search observation time period (Bhatia, para 0111 discloses the time taken to react or interact for users upon displaying contents (post search observation) “the user device 102/103 and/or meet-up interface 303 may include the timer that starts upon receiving the meet-up recommendation 108. In some embodiments, the predetermined period of time may include any suitable time period for user interaction with the meet-up interface 303”);
Moon further teaches receiving a search event data associated with the plurality of search results (Moon, para 0034 discloses retrieving plurality of search result data (possible medical indication) for selection “information associated with a patient, such as age, gender, and an order description, when presenting a health care services provider (“provider”) with a list of possible medical indications for an order. This can result in the provider needing to review numerous possible medical indications to select one that best supports a particular order”);
Bhatia teaches and generating the plurality of delayed engagement relevance scores based at least in part on the clinical event data and the search event data(Bhatia, para 0111 discloses the time taken to react or interact (immediate/delayed engagement) for users upon receiving a content is being measured “the user device 102/103 and/or meet-up interface 303 may include the timer that starts upon receiving the meet-up recommendation 108. In some embodiments, the predetermined period of time may include any suitable time period for user interaction with the meet-up interface 303”; and Moon in para 0034 discloses retrieving users’ clinical diagnostic test information).
Claim 12-14 are rejected under 35 U.S.C. 103 as being unpatentable over Moon, Changsung et al(PGPUB Document No. 20230170092), hereafter referred as to “Moon”, in view of Wang, Zhenrui et al (PGPUB Document No. 20220245162), hereafter, referred to as “Wang”, in view of Sheshansh, Satyam et al (PGPUB Document No. 20230368130), hereafter, referred to as “Sheshansh”, ”, in further view of Bhatia, Vidit et al (PGPUB Document No. 20200104395), hereafter, referred to as “Bhatia”, in further view of Nida, Dean et al (PGPUB Document No. 20200279641), hereafter, referred to as “Nida”.
Regarding Claim 12 (Previously Presented), Moon, Wang, Sheshansh and Bhatia teach all the limitations of claim 9 and Moon further teaches wherein operations further comprise: determining a clinical event data object associated with a search result data object of the plurality of search result data objects, wherein the search query data object is associated with a user profile data object(Moon, para 0034 discloses retrieving users’ clinical diagnostic test information with user’ profile data such as age, gender etc. “information associated with a patient, such as age, gender, and an order description, when presenting a health care services provider (“provider”) with a list of possible medical indications”);
But Moon, Wang, Sheshansh and Bhatia don’t explicitly teach generating a cost difference variable based at least in part on inputting the user profile to an event-true cost-estimation machine learning model and an event-false cost-estimation machine learning model associated with the clinical event data; and generating an outcome relevance score associated with the search result based at least in part on the cost difference variable.
However in the same field of endeavor of content ranking Nida teaches generating a cost difference variable based at least in part on inputting the user profile to an event-true cost-estimation machine learning model and an event-false cost-estimation machine learning model associated with the clinical event data (Nida, para 0076 discloses a model for predicting various clinical cost “module 850 depicts one or more visual timeline(s) of future clinical predictions, cost predictions, and so on which may be generated by one or more models in the modeling module”; para 0076 further discloses future cost estimation ); and generating an outcome relevance score associated with the search result based at least in part on the cost difference variable (Nida, 0082 further discloses measure/score for relevance is being generated “method 1200 predicts the most actionable display modules based at least on the user profile and the patient data. For example, method 1200 may predict, with the actionability prediction model 740, which display modules are most relevant or actionable based on the type of user, as well as the patient data”; para 0090 discloses prediction of future health care cost “In a second example of the method, the plurality of risk scores comprises one or more of a prediction of future healthcare costs”; where Gao, col 10:32-35 further discloses a measure of outcome relevance score “A measure of relevance of a content item to the user is based on a measure of quality of the content item for the user, which may be based on the content presented by the content item. Based on the measures of relevance”).
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 clinical cost prediction of Nida into content ranking of Moon, Wang, Sheshansh and Bhatia to produce an expected result of presenting contents to users based on their clinical cost. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the clinical assessment method using risk scores to prioritize individuals for healthcare intervention on time(Nida, abstract).
Regarding Claim 13(Previously Presented), Moon, Wang, Sheshansh, Bhatia and Nida teach all the limitations of claim 12 and Moon further teaches wherein the operations further comprise: identifying, from a plurality of user profiles and based at least in part on a probability matching machine learning model, a first probability-matched user profile subset that is associated with the clinical event data (Moon, para 0034 discloses retrieving users’ clinical diagnostic test information with user’ profile data such as age, gender etc. “information associated with a patient, such as age, gender, and an order description, when presenting a health care services provider (“provider”) with a list of possible medical indications”)
Nida teaches and a second probability-matched user profile subset that is not associated with the clinical event data(Nida, para 0043 discloses plurality of models are being used for generating scores such user profile data (not using any medical history of claim) “the plurality of models may be configured to generate scores for individuals based on medical claims data, demographic data, and so on, in a variety of categories such as risk, cost, predicted length of stay, and so on. The output of each model may therefore comprise a list of members with associated scores in a given category”);
and a second probability-matched user profile subset that is not associated with the clinical event data; and training the event-true cost-estimation machine learning model based at least in part on the first probability-matched user profile subset and the event-false cost-estimation machine learning model based at least in part on the second probability-matched user profile subset (Nida, para 0043 discloses plurality of models are being used for prediction using cost, user profile data or medical claim data “the plurality of models may be configured to generate scores for individuals based on medical claims data, demographic data, and so on, in a variety of categories such as risk, cost, predicted length of stay, and so on. The output of each model may therefore comprise a list of members with associated scores in a given category”; where para 0044 and para 0090 disclose models are being trained).
Regarding Claim 14(Currently Amended), Moon, Wang, Sheshansh, Bhatia and Nida teach all the limitations of claim 12 and Nida further teaches wherein the operations further comprise: training the probability matching machine learning model based at least in part on one or more user profiles that are associated with the clinical event data and one or more user profiles that are not associated with the clinical event data (Nida, para 0043 discloses plurality of models are being used for generating scores such user historical medical claim data (clinical), user profile data (not using any medical history of claim) “the plurality of models may be configured to generate scores for individuals based on medical claims data, demographic data, and so on, in a variety of categories such as risk, cost, predicted length of stay, and so on. The output of each model may therefore comprise a list of members with associated scores in a given category”).
Response to Arguments
I. 35 U.S.C §103
Applicant’s arguments filed on 2/2/2026 have been fully considered but are
moot because the independent claim 1, 15 and 20 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
/AMY NG/Supervisory Patent Examiner, Art Unit 2164
Prosecution Insights