Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Status
Claims 1-20 are pending.
Specification
The specification is objected to as failing to provide proper antecedent basis for the claimed subject matter. See 37 CFR 1.75(d)(1) and MPEP § 608.01(o). Correction of the following is required:
Claim 1 recites:
submitting to the neural network, the content embeddings nearest to the query embeddings [[and the query embeddings]]
Claim 1 recites performing a search for the stored content embeddings nearest to the query embeddings;
Claim 1 recites submitting the query to the neural network to generate query embeddings
Claim 1 recites submitting the content to a neural network to generate content embeddings; and
receiving and storing the content embeddings;
Claim 7 recites wherein the chat history corresponds to at least one chat within a chat service, wherein the chat service submits the content to the neural network and
submits the query to the neural network.
Claim 11 recites similar language.
Claim Objections
Claim 1 is objected to because of the following informalities:
Claim 1 recites submitting to the neural network, the content embeddings nearest to the query embeddings [[and the query embeddings]]. The above language should be deleted as indicated.
Claim 11 recites similar language.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1 is/are rejected under 35 U.S.C. 103 as being unpatentable over Harikumar (US 2022/0391633) in view of Faust (US 2024/0143668) in view of Lee (US 2023/0169795) in view of Bryan (US 2024/0273355) in view of Marwah (US 2025/0036878).
Examiner Note: Hereafter, above references will be entered as reference combination A.
Harikumar discloses:
a processor coupled to a computer memory having stored thereon instructions that cause
the processor to perform:
Harikumar [0106] The components of the instance extraction system 106 can include software, hardware, or both. For example, the components of the instance extraction system 106 can include one or more instructions stored on a computer-readable storage medium and executable by processors of one or more computing devices (e.g., the user client device 108). When executed by the one or more processors, the computer-executable instructions of the instance extraction system 106 can cause the computing devices to perform the object clustering methods described herein. Alternatively, the components of the instance extraction system 106 can comprise hardware, such as a special purpose processing device to perform a certain function or group of functions. Additionally, or alternatively, the components of the instance extraction system 106 can include a combination of computer-executable instructions and hardware.
processing content comprising:
submitting the content to a neural network to generate content embeddings;
Harikumar [0050] The series of acts 200 illustrated in FIG. 2 further includes the act 206 of generating content embeddings and color embeddings for the extracted objects. For instance, a content embedding can include a digital representation of content portrayed in a digital image (e.g., a digital representation of an identifying label or semantic information). In particular, a content embedding can comprise a low-dimensional vector that captures the semantic properties of an object portrayed within a digital image. In one or more embodiments, the instance extraction system 106 generates content embeddings that capture the semantic properties of the detected objects. In one example, the instance extraction system 106 utilizes a convolutional neural network to generate the content embeddings. For example, a content embedding can comprise a 2048-dimensional feature vector generated by a convolutional neural network.
receiving and storing the content embeddings;
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Faust discloses:
Faust claim 1, A computer-implemented method comprising: generating a plurality of approximate nearest neighbor (“aNN”) indexes, each aNN index storing content item embeddings for content items associated with one end-user of a feed of an online service;
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Faust for the purpose of providing a technique to facilitate filtering during candidate retrieval by an information retrieval system that utilizes embedding models, see [0001].
Furthermore, a skilled artisan would be motivated to look to the analogous art of Faust for teachings which are from the same field of endeavor as the claimed invention, i.e., content embeddings.
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Faust.
processing a query comprising:
submitting the query to the neural network to generate query embeddings; and
receiving the query embeddings;
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Lee discloses:
Lee [0005] Systems, methods, and software are described herein for automatically recognizing a cue (e.g., a gesture or a spoken word/sound combined with the gesture) within an input video by determining a part of an example video where the cue occurs, applying a feature of the part to a neural network (e.g., a few-shot learning model) to generate a positive embedding, applying a feature of each chunk of the input video to the neural network to generate a plurality of negative embeddings, applying a given one of the chunks to the neural network to generate a query embedding, and using the generated embeddings to determine whether the cue occurs in the given chunk. If the cue does not occur in the given chunk, the process may be repeated again for another chunk of the video until a determination of the cue is made, or the entire video has been processed and no cue has been determined.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Lee for the purpose of generating a positive embedding.
Furthermore, a skilled artisan would be motivated to look to the analogous art of Faust for teachings which are from the same field of endeavor as the claimed invention, i.e., query embeddings
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Lee.
performing a search for the stored content embeddings nearest to the query embeddings;
requesting the neural network to generate a query prompt, comprising: Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Faust discloses:
Faust [0023] Then, at inference time (e.g., when a query is being processed), the query embedding, corresponding with or representing the viewing end-user, is used in a similarity search to identify content item embeddings in the content item embedding space that are similar to the query embedding. For example, the query embedding may be used as an input to a “k” approximate nearest neighbor (“k-ANN”) algorithm to identify some number (“k”) of content items having content item embeddings that are similar to the query embedding. Here, similarity is represented as the distance between two embeddings. Typical measures of distance include the cosine distance or the inner product of the two vectors. Information retrieval systems that utilize embeddings in this manner may be referred to as embedding-based information retrieval systems.
submitting, to the neural network, the content embeddings nearest to the query
embeddings and the query embeddings; and
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Bryan discloses:
Bryan abstract Embodiments are disclosed for identifying matching content using neural content fingerprints. The method may include receiving a request to identify content matching a query content item, wherein the query content item is a time varying content item, generating, by an embedding network, a neural fingerprint for the query content item, identifying one or more candidate content items based on the neural fingerprint of the query content item, determining, by a ranking network, one or more similarity scores corresponding to the one or more candidate content items, and identifying one or more matching content items based on the one or more similarity scores.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Bryan for the purpose of receiving a request to identify content matching a query content item.
Furthermore, a skilled artisan would be motivated to look to the analogous art of Bryan for teachings which are from the same field of endeavor as the claimed invnetion, i.e., query content.
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Bryan.
receiving, from the neural network, the query prompt;
submitting the query prompt to the neural network; and
receiving a response to the query prompt from the neural network.
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Marwah discloses:
Marwah [0014] In some aspects, the techniques described herein relate to a method including: receiving a query from a user device; mapping the query to a latent semantic embedding space, modeling a number of document segments of a number of documents, wherein each of the number of document segments includes document content, and wherein a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content thereof; generating a prompt including the query and a set of nearest document segments in the latent semantic embedding space matching the query; submitting the prompt to a large language model (LLM) and receiving a response therefrom, wherein the response includes an answer to the query and indicia of a document segment of the number of document segments including document content matching the answer; and providing the response to the user device.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Marwah for the purpose of providing a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content.
Furthermore, a skilled artisan would be motivated to look to the analogous art of Marwah for teachings which are from the same field of endeavor as the claimed invention, i.e., submitting a prompt to a large language model (LLM) and receiving a response from the LLM.
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Marwah.
Claim(s) 2 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Kleindessner (US 12,229,179).
Reference combination A discloses elements of the claimed invention as noted but does not disclose wherein performing a search for the stored content embeddings nearest to the query embeddings comprises: performing a cosine similarity search for the stored content embeddings nearest to the query embeddings. However, Kleindessner discloses:
Kleindessner col 14, lines 30-40, Alternatively, the embedding data store 112 and/or user search service 116 may execute the similarity search using one of the cosine similarity search, approximate nearing neighbor (ANN) algorithms, k nearest neighbors (KNN) method, locality sensitive hashing (LSH), range queries, or any other vector clustering and/or similarity search algorithms to search media embedding(s) that are similar to the transformed query embedding.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify reference combination A to obtain above limitation based on the teachings of Kleindessner for the purpose of searching media embedding(s) that are similar to the transformed query embedding.
Claim(s) 3 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Kleindessner
Reference combination A in view of Kleindessner discloses wherein the stored content embeddings nearest to the query embeddings comprise next nearest neighbors.
Kleindessner col 14, lines 30-40, Alternatively, the embedding data store 112 and/or user search service 116 may execute the similarity search using one of the cosine similarity search, approximate nearing neighbor (ANN) algorithms, k nearest neighbors (KNN) method, locality sensitive hashing (LSH), range queries, or any other vector clustering and/or similarity search algorithms to search media embedding(s) that are similar to the transformed query embedding.
Claim(s) 4 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A.
Reference combination A discloses wherein processing the query further comprises: segmenting the query into text chunks; and submitting the text chunks to the neural network to generate query embeddings.
Lee [0005] Systems, methods, and software are described herein for automatically recognizing a cue (e.g., a gesture or a spoken word/sound combined with the gesture) within an input video by determining a part of an example video where the cue occurs, applying a feature of the part to a neural network (e.g., a few-shot learning model) to generate a positive embedding, applying a feature of each chunk of the input video to the neural network to generate a plurality of negative embeddings, applying a given one of the chunks to the neural network to generate a query embedding, and using the generated embeddings to determine whether the cue occurs in the given chunk. If the cue does not occur in the given chunk, the process may be repeated again for another chunk of the video until a determination of the cue is made, or the entire video has been processed and no cue has been determined.
Claim(s) 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Wagh (US 2025/0117480).
Reference combination A discloses elements of the claimed invention as noted but does not disclose wherein the response to the query prompt is human readable.
However, Wagh discloses:
Wagh [0040] In some embodiments, machine learning model(s) may comprise one or more large language models (LLMs). A LLM deployed as all or part of machine learning model(s) 120 may be configured to accept human-readable and/or machine-readable text-based input, code input, tabular input, graphical input, and/or input in any one or more other suitable data formats or file formats. The LLM may be configured to process the received data and to generate output data (including privacy constraint compliance determination data, explainability data, and/or data requesting further information and/or human input) in any suitable format. The format of output data generated by the LLM may include human-readable and/or machine-readable text, tabular output, graphical output, and/or output in any one or more other suitable data formats or file formats.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify reference combination A to obtain above limitation based on the teachings of Wagh for the purpose of configuring a LLM to process received data and to generate output data.
Claim(s) 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Murakhovska (US 2024/0428068).
Reference combination A discloses elements of the claimed invention as noted but does not disclose further comprising: receiving a chat history, wherein requesting the neural network to generate a query prompt further comprises: submitting the chat history to the neural network. However, Murakhovska discloses:
Murakhovska [0047] For example, when a knowledge search module 310 is determined, the module 310 may educate a buyer (e.g., a simulated Shopper bot 114 or a human shopper) by incorporating expert domain knowledge into the conversation, which comprises: 1) query generation 322, and 2) retrieval 320 from a knowledge article database 319. Specifically, an LLM may be used to generate a query based on the chat history 202. A FAISS retriever 320 may be used to lookup relevant knowledge article paragraphs. For example, top three paragraphs may be concatenated (separated by “\n\n”) and fed as external knowledge to the Response Generation module 330.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify reference combination A to obtain above limitation based on the teachings of Murakhovska for the purpose of educating a buyer (e.g., a simulated Shopper bot 114 or a human shopper) by incorporating expert domain knowledge into the conversation.
Claim(s) 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Murakhovska
Reference combination A in view of Murakhovska discloses wherein the chat history corresponds to at least one chat within a chat service, wherein the chat service submits the content to the neural network and submits the query to the neural network.
Marwah [0014] In some aspects, the techniques described herein relate to a method including: receiving a query from a user device; mapping the query to a latent semantic embedding space, modeling a number of document segments of a number of documents, wherein each of the number of document segments includes document content, and wherein a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content thereof; generating a prompt including the query and a set of nearest document segments in the latent semantic embedding space matching the query; submitting the prompt to a large language model (LLM) and receiving a response therefrom, wherein the response includes an answer to the query and indicia of a document segment of the number of document segments including document content matching the answer; and providing the response to the user device.
Claim(s) 8 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A.
Reference combination A discloses further comprising: receiving content selections, wherein submitting the content to the neural network comprises submitting the content selections.
Harikumar [0050] The series of acts 200 illustrated in FIG. 2 further includes the act 206 of generating content embeddings and color embeddings for the extracted objects. For instance, a content embedding can include a digital representation of content portrayed in a digital image (e.g., a digital representation of an identifying label or semantic information). In particular, a content embedding can comprise a low-dimensional vector that captures the semantic properties of an object portrayed within a digital image. In one or more embodiments, the instance extraction system 106 generates content embeddings that capture the semantic properties of the detected objects. In one example, the instance extraction system 106 utilizes a convolutional neural network to generate the content embeddings. For example, a content embedding can comprise a 2048-dimensional feature vector generated by a convolutional neural network.
Claim(s) 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of non-functional descriptive material.
Reference combination A discloses elements of the claimed invention as noted but does not disclose wherein the content selections comprise documents, videos, audio, images, application data files, or a combination thereof.
Above limitation is rejected on the basis that no new or nonobvious functional relationship exists with the known method of generating context for a query from the data source [0006]
Claim(s) 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A.
Reference combination A discloses wherein the content selections are received from a content repository.
Marwah [0014] In some aspects, the techniques described herein relate to a method including: receiving a query from a user device; mapping the query to a latent semantic embedding space, modeling a number of document segments of a number of documents, wherein each of the number of document segments includes document content, and wherein a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content thereof; generating a prompt including the query and a set of nearest document segments in the latent semantic embedding space matching the query; submitting the prompt to a large language model (LLM) and receiving a response therefrom, wherein the response includes an answer to the query and indicia of a document segment of the number of document segments including document content matching the answer; and providing the response to the user device.
Claim(s) 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Harikumar (US 2022/0391633) in view of Faust (US 2024/0143668) in view of Lee (US 2023/0169795) in view of Bryan (US 2024/0273355) in view of Marwah (US 2025/0036878).
Examiner Note: Hereafter, above references will be entered as reference combination A.
Harikumar discloses:
a processor coupled to a computer memory having stored thereon instructions that cause
the processor to perform:
Harikumar [0106] The components of the instance extraction system 106 can include software, hardware, or both. For example, the components of the instance extraction system 106 can include one or more instructions stored on a computer-readable storage medium and executable by processors of one or more computing devices (e.g., the user client device 108). When executed by the one or more processors, the computer-executable instructions of the instance extraction system 106 can cause the computing devices to perform the object clustering methods described herein. Alternatively, the components of the instance extraction system 106 can comprise hardware, such as a special purpose processing device to perform a certain function or group of functions. Additionally, or alternatively, the components of the instance extraction system 106 can include a combination of computer-executable instructions and hardware.
processing content comprising:
submitting the content to a neural network to generate content embeddings;
Harikumar [0050] The series of acts 200 illustrated in FIG. 2 further includes the act 206 of generating content embeddings and color embeddings for the extracted objects. For instance, a content embedding can include a digital representation of content portrayed in a digital image (e.g., a digital representation of an identifying label or semantic information). In particular, a content embedding can comprise a low-dimensional vector that captures the semantic properties of an object portrayed within a digital image. In one or more embodiments, the instance extraction system 106 generates content embeddings that capture the semantic properties of the detected objects. In one example, the instance extraction system 106 utilizes a convolutional neural network to generate the content embeddings. For example, a content embedding can comprise a 2048-dimensional feature vector generated by a convolutional neural network.
receiving and storing the content embeddings;
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Faust discloses:
Faust claim 1, A computer-implemented method comprising: generating a plurality of approximate nearest neighbor (“aNN”) indexes, each aNN index storing content item embeddings for content items associated with one end-user of a feed of an online service;
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Faust for the purpose of providing a technique to facilitate filtering during candidate retrieval by an information retrieval system that utilizes embedding models, see [0001].
Furthermore, a skilled artisan would be motivated to look to the analogous art of Faust for teachings which are from the same field of endeavor as the claimed invention, i.e., content embeddings.
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Faust.
processing a query comprising:
submitting the query to the neural network to generate query embeddings; and
receiving the query embeddings;
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Lee discloses:
Lee [0005] Systems, methods, and software are described herein for automatically recognizing a cue (e.g., a gesture or a spoken word/sound combined with the gesture) within an input video by determining a part of an example video where the cue occurs, applying a feature of the part to a neural network (e.g., a few-shot learning model) to generate a positive embedding, applying a feature of each chunk of the input video to the neural network to generate a plurality of negative embeddings, applying a given one of the chunks to the neural network to generate a query embedding, and using the generated embeddings to determine whether the cue occurs in the given chunk. If the cue does not occur in the given chunk, the process may be repeated again for another chunk of the video until a determination of the cue is made, or the entire video has been processed and no cue has been determined.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Lee for the purpose of generating a positive embedding.
Furthermore, a skilled artisan would be motivated to look to the analogous art of Faust for teachings which are from the same field of endeavor as the claimed invention, i.e., query embeddings
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Lee.
performing a search for the stored content embeddings nearest to the query embeddings;
requesting the neural network to generate a query prompt, comprising: Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Faust discloses:
Faust [0023] Then, at inference time (e.g., when a query is being processed), the query embedding, corresponding with or representing the viewing end-user, is used in a similarity search to identify content item embeddings in the content item embedding space that are similar to the query embedding. For example, the query embedding may be used as an input to a “k” approximate nearest neighbor (“k-ANN”) algorithm to identify some number (“k”) of content items having content item embeddings that are similar to the query embedding. Here, similarity is represented as the distance between two embeddings. Typical measures of distance include the cosine distance or the inner product of the two vectors. Information retrieval systems that utilize embeddings in this manner may be referred to as embedding-based information retrieval systems.
submitting, to the neural network, the content embeddings nearest to the query
embeddings and the query embeddings; and
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Bryan discloses:
Bryan abstract Embodiments are disclosed for identifying matching content using neural content fingerprints. The method may include receiving a request to identify content matching a query content item, wherein the query content item is a time varying content item, generating, by an embedding network, a neural fingerprint for the query content item, identifying one or more candidate content items based on the neural fingerprint of the query content item, determining, by a ranking network, one or more similarity scores corresponding to the one or more candidate content items, and identifying one or more matching content items based on the one or more similarity scores.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Bryan for the purpose of receiving a request to identify content matching a query content item.
Furthermore, a skilled artisan would be motivated to look to the analogous art of Bryan for teachings which are from the same field of endeavor as the claimed invnetion, i.e., query content.
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Bryan.
receiving, from the neural network, the query prompt;
submitting the query prompt to the neural network; and
receiving a response to the query prompt from the neural network.
Harikumar discloses elements of the claimed invention as noted but does not disclose above limitation. However, Marwah discloses:
Marwah [0014] In some aspects, the techniques described herein relate to a method including: receiving a query from a user device; mapping the query to a latent semantic embedding space, modeling a number of document segments of a number of documents, wherein each of the number of document segments includes document content, and wherein a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content thereof; generating a prompt including the query and a set of nearest document segments in the latent semantic embedding space matching the query; submitting the prompt to a large language model (LLM) and receiving a response therefrom, wherein the response includes an answer to the query and indicia of a document segment of the number of document segments including document content matching the answer; and providing the response to the user device.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Harikumar to obtain above limitation based on the teachings of Marwah for the purpose of providing a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content.
Furthermore, a skilled artisan would be motivated to look to the analogous art of Marwah for teachings which are from the same field of endeavor as the claimed invention, i.e., submitting a prompt to a large language model (LLM) and receiving a response from the LLM.
Based on the above, a skilled artisan would reasonably expect success when combining Harikumar and Marwah.
Claim(s) 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Kleindessner (US 12,229,179).
Reference combination A discloses elements of the claimed invention as noted but does not disclose wherein performing a search for the stored content embeddings nearest to the query embeddings comprises: performing a cosine similarity search for the stored content embeddings nearest to the query embeddings. However, Kleindessner discloses:
Kleindessner col 14, lines 30-40, Alternatively, the embedding data store 112 and/or user search service 116 may execute the similarity search using one of the cosine similarity search, approximate nearing neighbor (ANN) algorithms, k nearest neighbors (KNN) method, locality sensitive hashing (LSH), range queries, or any other vector clustering and/or similarity search algorithms to search media embedding(s) that are similar to the transformed query embedding.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify reference combination A to obtain above limitation based on the teachings of Kleindessner for the purpose of searching media embedding(s) that are similar to the transformed query embedding.
Claim(s) 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Kleindessner
Reference combination A in view of Kleindessner discloses wherein the stored content embeddings nearest to the query embeddings comprise next nearest neighbors.
Kleindessner col 14, lines 30-40, Alternatively, the embedding data store 112 and/or user search service 116 may execute the similarity search using one of the cosine similarity search, approximate nearing neighbor (ANN) algorithms, k nearest neighbors (KNN) method, locality sensitive hashing (LSH), range queries, or any other vector clustering and/or similarity search algorithms to search media embedding(s) that are similar to the transformed query embedding.
Claim(s) 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A.
Reference combination A discloses wherein processing the query further comprises: segmenting the query into text chunks; and submitting the text chunks to the neural network to generate query embeddings.
Lee [0005] Systems, methods, and software are described herein for automatically recognizing a cue (e.g., a gesture or a spoken word/sound combined with the gesture) within an input video by determining a part of an example video where the cue occurs, applying a feature of the part to a neural network (e.g., a few-shot learning model) to generate a positive embedding, applying a feature of each chunk of the input video to the neural network to generate a plurality of negative embeddings, applying a given one of the chunks to the neural network to generate a query embedding, and using the generated embeddings to determine whether the cue occurs in the given chunk. If the cue does not occur in the given chunk, the process may be repeated again for another chunk of the video until a determination of the cue is made, or the entire video has been processed and no cue has been determined.
Claim(s) 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Wagh (US 2025/0117480).
Reference combination A discloses elements of the claimed invention as noted but does not disclose wherein the response to the query prompt is human readable.
However, Wagh discloses:
Wagh [0040] In some embodiments, machine learning model(s) may comprise one or more large language models (LLMs). A LLM deployed as all or part of machine learning model(s) 120 may be configured to accept human-readable and/or machine-readable text-based input, code input, tabular input, graphical input, and/or input in any one or more other suitable data formats or file formats. The LLM may be configured to process the received data and to generate output data (including privacy constraint compliance determination data, explainability data, and/or data requesting further information and/or human input) in any suitable format. The format of output data generated by the LLM may include human-readable and/or machine-readable text, tabular output, graphical output, and/or output in any one or more other suitable data formats or file formats.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify reference combination A to obtain above limitation based on the teachings of Wagh for the purpose of configuring a LLM to process received data and to generate output data.
Claim(s) 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Murakhovska (US 2024/0428068).
Reference combination A discloses elements of the claimed invention as noted but does not disclose further comprising: receiving a chat history, wherein requesting the neural network to generate a query prompt further comprises: submitting the chat history to the neural network. However, Murakhovska discloses:
Murakhovska [0047] For example, when a knowledge search module 310 is determined, the module 310 may educate a buyer (e.g., a simulated Shopper bot 114 or a human shopper) by incorporating expert domain knowledge into the conversation, which comprises: 1) query generation 322, and 2) retrieval 320 from a knowledge article database 319. Specifically, an LLM may be used to generate a query based on the chat history 202. A FAISS retriever 320 may be used to lookup relevant knowledge article paragraphs. For example, top three paragraphs may be concatenated (separated by “\n\n”) and fed as external knowledge to the Response Generation module 330.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify reference combination A to obtain above limitation based on the teachings of Murakhovska for the purpose of educating a buyer (e.g., a simulated Shopper bot 114 or a human shopper) by incorporating expert domain knowledge into the conversation.
Claim(s) 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of Murakhovska
Reference combination A in view of Murakhovska discloses wherein the chat history corresponds to at least one chat within a chat service, wherein the chat service submits the content to the neural network and submits the query to the neural network.
Marwah [0014] In some aspects, the techniques described herein relate to a method including: receiving a query from a user device; mapping the query to a latent semantic embedding space, modeling a number of document segments of a number of documents, wherein each of the number of document segments includes document content, and wherein a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content thereof; generating a prompt including the query and a set of nearest document segments in the latent semantic embedding space matching the query; submitting the prompt to a large language model (LLM) and receiving a response therefrom, wherein the response includes an answer to the query and indicia of a document segment of the number of document segments including document content matching the answer; and providing the response to the user device.
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A.
Reference combination A discloses further comprising: receiving content selections, wherein submitting the content to the neural network comprises submitting the content selections.
Harikumar [0050] The series of acts 200 illustrated in FIG. 2 further includes the act 206 of generating content embeddings and color embeddings for the extracted objects. For instance, a content embedding can include a digital representation of content portrayed in a digital image (e.g., a digital representation of an identifying label or semantic information). In particular, a content embedding can comprise a low-dimensional vector that captures the semantic properties of an object portrayed within a digital image. In one or more embodiments, the instance extraction system 106 generates content embeddings that capture the semantic properties of the detected objects. In one example, the instance extraction system 106 utilizes a convolutional neural network to generate the content embeddings. For example, a content embedding can comprise a 2048-dimensional feature vector generated by a convolutional neural network.
Claim(s) 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A in view of non-functional descriptive material.
Reference combination A discloses elements of the claimed invention as noted but does not disclose wherein the content selections comprise documents, videos, audio, images, application data files, or a combination thereof.
Above limitation is rejected on the basis that no new or nonobvious functional relationship exists with the known method of generating context for a query from the data source [0006]
Claim(s) 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over reference combination A.
Reference combination A discloses wherein the content selections are received from a content repository.
Marwah [0014] In some aspects, the techniques described herein relate to a method including: receiving a query from a user device; mapping the query to a latent semantic embedding space, modeling a number of document segments of a number of documents, wherein each of the number of document segments includes document content, and wherein a distance between any two of the number of document segments in the latent semantic embedding space is proportional to a degree of similarity between the document content thereof; generating a prompt including the query and a set of nearest document segments in the latent semantic embedding space matching the query; submitting the prompt to a large language model (LLM) and receiving a response therefrom, wherein the response includes an answer to the query and indicia of a document segment of the number of document segments including document content matching the answer; and providing the response to the user device.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ETIENNE PIERRE LEROUX whose telephone number is (571)272-4022. The examiner can normally be reached M-F 8:00 am to 4:30 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, Apu Mofiz can be reached at 571 272 4080. 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.
/ETIENNE P LEROUX/Primary Examiner of Art Unit 2161