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 .
This action is responsive to the filing on 12/30/2024. Claim(s) 1 and 20 have been
amended. Claims 1-30 are pending in this case. Claims 1 and 10 are independent claims.
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.
Claims 1-6, 9-15, 17, 20-22, 23-24 and 26-30 are rejected under 35 U.S.C 103 as being unpatentable over Aravamudan (US Patent No.12,211,598 B1), hereafter referred to as Aravamudan in view of LIU et al. (US Pub No.: 20200065619 A1), hereafter referred to as LIU and further in view of SONG et al. (US Pub No.: 20240273369 A1), hereafter referred to as SONG.
With respect to claim 1, Aravamudan disclose:
A computer system, the computer system comprising: a network interface (In Col. 3, lines 9–14, Aravamudan discloses that a computing device can connect and communicate with other devices through a network interface device. This device helps connect the computing device to different types of networks and devices.)
at least one processing device comprising one or more arithmetic logic units, registers, and buses, the at least one processing device operable to: detect a prompt entered into a user interface field from a user device provided to a generative model (In FIG. 1 and Col. 10-11, lines 66–25, Aravamudan disclosed, a computing device may configure generative machine learning models to analyze input data to one or more predefined templates, computing device configured to pinpoint specific errors in prompt and/or electronic medical record EMR In Col. 66, lines 16-23, Aravamudan disclose The computer system has a processor 904 and a memory that talk to each other and other parts through a bus. The bus can be different types, like a memory bus, a memory controller, a peripheral bus, a local bus, or a mix of these, using different bus designs. In Col. 66, lines 24–27, disclose a processor incorporating logical circuitry for performing arithmetic and logical operations, such as an arithmetic and logic unit (ALU).)
detect a response to the prompt, output by the generative model (In Col. 11, lines 3–16, Aravamudan discloses the computing device configured to pinpoint specific errors in the prompt. The computing device configured to implement generative machine learning models to incorporate additional models to detect additional instances of a prompt. In some cases, errors may be classified into different categories or severity levels and the generative machine learning model, such as, without limitation, GAN, is configured to generate first generative model output.)
estimate a contribution of a first item of content, used to train the generative model, to the generative model output, wherein the estimated contribution is determined via an ensemble of attribution techniques comprising two or more of: a usefulness projection with respect to a first domain, the usefulness projected based at least in part on descriptive and/or source labels, wherein the descriptive and/or source labels are associated with the first item of content (Examiner selects: (a)“ a usefulness projection with respect to a first domain… as well as (c)“generate prompts comprising content from a plurality of data sources including the first item of content…” In Col. 10, lines 12–24, Aravamudan discloses that a GAN is configured to receive a prompt and/or EMR as input and generate the corresponding first-generative model output containing information describing or evaluating the performance of one or more instances of prompt and/or EMR. The discriminator of GAN may evaluate the authenticity of the generated content by comparing it to real first generative model output 160, for example, discriminator may distinguish between genuine and generated content and providing feedback to generator to improve the model performance. In Col. 9, lines 25–45, Aravamudan discloses that one or more generative machine learning models containing Naïve Bayes classifiers may be trained on labeled training data, estimating conditional probabilities using techniques such as Maximum Likelihood Estimation (MLE).)
a determined modification to neuron weights of the generative model, wherein when the first item of content is used to train the generative model modification of the neuron weights are monitored during training and corresponding data is stored
generate prompts comprising content from a plurality of data sources including the first item of content, provide the generated prompts to the generative model to generate respective outputs, and determine similarities between the generative model outputs and respective content from the content sources (In Fig. 8 and Col. 28-35, Aravamudan discloses creating a prompt based on the input from the user. In other cases, the prompt is made by putting the first user input into a specific template; and the EMR database query is made by using the first user input in another template.)
a stylometric analysis of the generative model output performed using a Support Vector machine configured to perform authorship attribution by finding a hyperplane that best separates different classes in a feature space, a random forest, and/or neural network, wherein at least two of the attribution techniques are weighted differently in estimating the contribution of the first item of content to the generative model output
transfer the first token amount to a first destination transmit, using the network interface, the feedback comprising the first token amount generated determined based at least in part on the estimated contribution of the first item of content, used to train the generative model, to the generative model output to one or more networked destinations (In Fig. 8 and Col. 64, lines 43–46, Aravamudan discloses using a conversational interface displayed using the user interface, displaying the first generative model output to the user.)
With respect to claim 1, Aravamudan does not explicitly disclose:
wherein at least two of the attribution techniques are weighted differently in estimating the contribution of the first item of content to the generative model output
determine a feedback comprising a first token amount based at least in part on the estimated contribution of the first item of content, used to train the generative model, to the generative model output
However, it is known by LIU to disclose:
Wherein at least two of the attribution techniques are weighted differently in estimating the contribution of the first item of content to the generative model output (In Fig. 5 and paragraph [0056-0057], LIU discloses two noise images: the first one, N.sub.1, and the second one, N.sub.2. These images are different from each other. Each noise image can have M channels, where M is a positive whole number, starting from 1. For example, in this case, M can be 1. The noise image is used together with the first training image to help the generative neural network create more details during image conversion. In other cases, M can also be 3, meaning we add information from 3 channels of the noise image to the 3 color channels of the first training image, and then this updated first training image is sent to the generative neural network for conversion.)
Aravamudan and LIU are analogous pieces of art because both references concern using a generative neural network. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Aravamudan, with generating a first generative model output as a function of the prompt and the EMR using a trained generative machine learning model as taught by Aravamudan, with input of the generative neural network includes a noise image channel and N channels of the input image as taught by LIU. The motivation for doing so would have been to improve the model performance (See(Col. 10, lines 12-24))
With respect to claim 1, Aravamudan in view of LIU, does not explicitly disclose:
Determine a feedback comprising a first token amount based at least in part on the estimated contribution of the first item of content, used to train the generative model, to the generative model output
However, it is known by SONG to disclose:
Determine a feedback comprising a first token amount based at least in part on the estimated contribution of the first item of content, used to train the generative model, to the generative model output (In Fig. 4 and paragraph [0072], SONG discloses determining quality scores for each candidate's reply using the discriminator scores. The quality score for a candidate's reply may be calculated by adding up the discriminator scores for that reply, with some scores given more importance than others. In paragraph [0081], the cross-entropy-based objective Ice may determine token-level cross-entropy loss based on human annotated replies.)
Aravamudan in view of LIU and SONG are analogous pieces of art because both references concern using a generative neural network. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify SONG, with determining quality scores for the candidate replies based on the discriminator scores as taught by SONG. The motivation for doing so would have been to improve operating efficiency and processing rate of the network (See[0051] of LIU)
Regarding claim 2, Aravamudan in view of LIU and SONG discloses element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the system is configured to estimate contribution percentages of a plurality of content items to the generative model output and to transmit corresponding pro rata feedback to respective sources of items in the plurality of content items (In Col. 10, lines 15–24, Aravamudan discloses that GAN may be configured to receive prompt and/or EMR as input and generates the corresponding first-generative model output containing information describing or evaluating the performance of one or more instances of prompt and/or EMR. On the other hand, discriminator of GAN may evaluate the authenticity of the generated content by comparing it to real first generative model output, for example, discriminator may distinguish between genuine and generated content and providing feedback to generator to improve the model performance.)
Regarding claim 3, Aravamudan in view of LIU and SONG discloses element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the generative model comprises a large language model (In Col. 7-8, lines 67–3, Aravamudan discloses a generative machine learning model including a large language model.)
Regarding claim 4, Aravamudan in view of LIU and SONG discloses element of claim 1. Aravamudan also discloses:
The computer system as defined in Claim 1, wherein the generative model comprises an image generator (In Col. 39, lines 21–28, Aravamudan discloses a generative model with a generator that creates new images.)
Regarding claim 5, Aravamudan in view of LIU and SONG discloses element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the generative model output comprises text, still image data, video image data and/or audio data (In Col.8, lines 13–16, Aravamudan discloses the generative model output comprises text, image, video, audio.)
Regarding claim 6, Aravamudan in view of LIU and SONG discloses element of claim 1. Aravamudan also discloses:
The computer system as defined in Claim 1, wherein the first item of content comprises still image data, video image data, and/or audio data (In Col. 8, lines 23–26, Aravamudan discloses the first generative model output and/or the like in any data structure as described herein (e.g., text, image, video, audio).)
Regarding claim 9, Aravamudan in view of LIU and SONG discloses element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the estimated contribution of the first item of content to the generative model output comprises an estimated contribution of vocabulary choice, sentence structure, grammar and punctuation, tone and voice, themes and topics, and/or rhetorical devices to the generative model output (In Col. 41, lines 28–33, Aravamudan includes applying voice conversion techniques to alter the patient's voice in the audio records, replacing it with a synthetic voice, generated using generative model 228 that maintains the linguistic content but removes the identifiable voice characteristics and/or adds relative context.)
Regarding claim 10, Aravamudan in view of LIU and SONG discloses element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the estimated contribution of the first item of content to the generative model output comprises an estimated contribution of symbols, shapes, motifs, and/or iconography to the generative model output (In Col. 41, lines 40–51, Aravamudan discloses any symbols usable as textual data.)
Regarding claim 11, Aravamudan in view of LIU and SONG discloses element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the estimated contribution of the first item of content to the generative model output comprises an estimated contribution to one or more claims of the generative model output (In Col. 8, lines 26–29, Aravamudan discloses that the machine learning module described herein may generate one or more generative machine learning models that are trained on one or more sets of training data. )
Regarding claim 12, Aravamudan in view of LIU and SONG discloses the element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein transmitting the feedback generated based at least in part on the estimated contribution of the first item of content to the generative model output to one or more networked destinations, further comprises transmitting feedback to a plurality of networked destinations based at least in part on estimated percentage contributions of a plurality of items of content to the generative model output (In Col. 61, lines 8–14, Aravamudan discloses that a plurality of user feedback scores may be averaged to determine an accuracy score. In some embodiments, a cohort accuracy score may be determined for particular cohorts of persons. For example, user feedback for users belonging to a particular cohort of persons may be averaged together to determine the cohort accuracy score for that particular cohort of persons and used as described above.)
Regarding claim 13, Aravamudan in view of LIU and SONG discloses element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the system is configured to transmit an aggregated feedback for a first period of time to the one or more networked destinations (In Col. 10, lines 20–24, Aravamudan discloses that GAN may evaluate the authenticity of the generated content by comparing it to real first generative model output. For example, a discriminator may distinguish between genuine and generated content and provide feedback to the generator to improve the model performance. )
Regarding claim 14, Aravamudan in view of LIU and SONG discloses the element of claim 1. In addition, Aravamudan disclose:
The computer system as defined in Claim 1, wherein the system is operable to estimate the contribution of the first item of content, used to train the generative model, to the generative model output based at least in part on labels associated with the first item of content, changes in weights of the generative model caused at least partly by training of the generative model using the first item of content, and/or based on an analysis of the output of the generative model (In Col. 24, lines 53–60, Aravamudan discloses the application of a back propagation algorithm may involve computing a gradient of a loss function based on the weights of a neural network, and modifying the weights based on the gradient. Machine learning models and neural networks )
Regarding claim 15, Aravamudan in view of LIU and SONG discloses element of claim 1. Aravamudan disclose:
The computer system as defined in Claim 1, wherein the feedback comprises label weights associated with labels assigned to the first item of content, an identification of a contribution to an adjustment of a weight of the generative model, and/or a token (In Col. 56, lines 19–21, Aravamudan discloses adjusting the connections and weights between nodes in adjacent layers of the neural network to produce the desired values at the output nodes.)
Regarding claim 17, Aravamudan in view of LIU and SONG discloses element of claim 1. Aravamudan disclose:
The computer system as defined in Claim 1, wherein the computer system is operable to: chunk text of at least a first document into a plurality of overlapping chunks (In Col. 30, lines 30–35, Aravamudan discloses a plurality of text documents, wherein each document of the plurality of text documents may be transformed into a vector using TF-IDF or word embeddings.)
generate embeddings comprising vectors corresponding to the plurality of overlapping chunks (In Col. 30, lines 35–38, Aravamudan discloses cases, each dimension in each vector may represent, for example, the significance of a word or phrase within the corresponding document in the context of the entire database.)
and store the embeddings corresponding to the plurality of overlapping chunks in a vector database (In Col. 30-31, lines 67-3, Aravamudan discloses that a processor 204 may be configured to perform research, classification, topic modeling, content generation, and/or the like on such text data stored in database 212.)
With respect to claim 20, Aravamudan disclose:
A computer-implemented method, the method comprising: accessing an item of generated content from non-transitory memory: (In Col. 28, lines 38- 46, Aravamudan disclose apparatus include processor and/or memory. Processor may be configured to access a database containing a plurality of private data elements belonging to at least a private record. A database may include a collection of data that can be accessed, managed, and updated.)
estimate a contribution of a first item of content, used to train the generative model, to the generative model output, wherein the estimated contribution is determined via an ensemble of attribution techniques comprising two or more of: a usefulness projection with respect to a first domain, the usefulness projected based at least in part on descriptive and/or source labels, wherein the descriptive and/or source labels are associated with the first item of content (Examiner selects: (a)“ a usefulness projection with respect to a first domain… as well as (c)“generate prompts comprising content from a plurality of data sources including the first item of content…” In Col. 10, lines 12–24, Aravamudan discloses that a GAN is configured to receive a prompt and/or EMR as input and generate the corresponding first-generative model output containing information describing or evaluating the performance of one or more instances of prompt and/or EMR. The discriminator of GAN may evaluate the authenticity of the generated content by comparing it to real first generative model output 160, for example, discriminator may distinguish between genuine and generated content and providing feedback to generator to improve the model performance. In Col. 9, lines 25–45, Aravamudan discloses that one or more generative machine learning models containing Naïve Bayes classifiers may be trained on labeled training data, estimating conditional probabilities using techniques such as Maximum Likelihood Estimation (MLE).)
a determined modification to neuron weights of the generative model, wherein when the first item of content is used to train the generative model modification of the neuron weights are monitored during training and corresponding data is store
generate prompts comprising content from a plurality of data sources including the first item of content, provide the generated prompts to the generative model to generate respective outputs, and determine similarities between the generative model outputs and respective content from the content sources (In Fig. 8 and Col. 28-35, Aravamudan discloses creating a prompt based on the input from the user. In other cases, the prompt is made by putting the first user input into a specific template; and the EMR database query is made by using the first user input in another template.)
a stylometric analysis of the generative model output performed using a Support Vector machine configured to perform authorship attribution by finding a hyperplane that best separates different classes in a feature space, a random forest, and/or neural network
transferring the first token amount to a first destination, the feedback comprising the first token amount determined based at least in part on the estimated contribution of the first item of content to the generated content to one or more networked destinations (In Fig. 8 and Col. 64, lines 43–46, Aravamudan discloses using a conversational interface displayed using the user interface, displaying the first generative model output to the user.)
With respect to claim 20, Aravamudan does not explicitly disclose:
wherein at least two of the attribution techniques are weighted differently in estimating the contribution of the first item of content to the generative model output
determining a feedback comprising a first token amount based at least in part on the estimated contribution of the first item of content to the generated content
However, it is known by LIU to disclose:
Wherein at least two of the attribution techniques are weighted differently in estimating the contribution of the first item of content to the generative model output (In Fig. 5 and paragraph [0056-0057], LIU discloses two noise images: the first one, N.sub.1, and the second one, N.sub.2. These images are different from each other. Each noise image can have M channels, where M is a positive whole number, starting from 1. For example, in this case, M can be 1. The noise image is used together with the first training image to help the generative neural network create more details during image conversion. In other cases, M can also be 3, meaning we add information from 3 channels of the noise image to the 3 color channels of the first training image, and then this updated first training image is sent to the generative neural network for conversion.)
Aravamudan and LIU are analogous pieces of art because both references concern using a generative neural network. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Aravamudan, with generating a first generative model output as a function of the prompt and the EMR using a trained generative machine learning model as taught by Aravamudan, with input of the generative neural network includes a noise image channel and N channels of the input image as taught by LIU. The motivation for doing so would have been to improve the model performance (See(Col. 10, lines 12-24))
With respect to claim 20, Aravamudan and LIU does not explicitly disclose:
determining a feedback comprising a first token amount based at least in part on the estimated contribution of the first item of content to the generated content
However, it is known by SONG to disclose:
Determining a feedback comprising a first token amount based at least in part on the estimated contribution of the first item of content to the generated content (In Fig. 4 and paragraph [0072], SONG discloses determining quality scores for each candidate's reply using the discriminator scores. The quality score for a candidate's reply may be calculated by adding up the discriminator scores for that reply, with some scores given more importance than others. In paragraph [0081], the cross-entropy-based objective Ice may determine token-level cross-entropy loss based on human annotated replies.)
Aravamudan in view of LIU and SONG are analogous pieces of art because both references concern using a generative neural network. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify SONG, with determining quality scores for the candidate replies based on the discriminator scores as taught by SONG. The motivation for doing so would have been to improve operating efficiency and processing rate of the network (See[0051] of LIU)
Regarding claim 21, Aravamudan in view of LIU and SONG discloses element of claim 20. In addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, the method further comprising: estimating contribution percentages of a plurality of content items to the generated content; and transmitting corresponding pro rata feedback to respective sources of items in the plurality of content items (In Col. 10, lines 15–24, Aravamudan discloses that GAN may be configured to receive prompt and/or EMR as input and generates the corresponding first-generative model output containing information describing or evaluating the performance of one or more instances of prompt and/or EMR. On the other hand, discriminator of GAN may evaluate the authenticity of the generated content by comparing it to real first generative model output, for example, discriminator may distinguish between genuine and generated content and providing feedback to generator to improve the model performance.)
Regarding claim 22, Aravamudan in view of LIU and SONG discloses element of claim 20. In addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, wherein the generated content comprises text, image data, or audio data (In Col.8, lines 13–16, Aravamudan discloses the generative model output comprises text, image, video, audio.)
Regarding claim 23, Aravamudan in view of LIU and SONG discloses element of claim 20. In addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, wherein the generated content is generated using a generative model, wherein estimating the contribution of the first item of content to the generated content is based at least in part on labels associated with the first item of content, changes in weights of a generative model used to generate the generated content caused at least partly by training of the generative model using the first item of content, and/or based on an analysis of the output of the generative model (In Col. 24, lines 53–60, Aravamudan discloses the application of a back propagation algorithm may involve computing a gradient of a loss function based on the weights of a neural network, and modifying the weights based on the gradient. Machine learning models and neural networks )
Regarding claim 24, Aravamudan in view of LIU and SONG discloses element of claim 20. Aravamudan also discloses:
The computer-implemented method as defined in Claim 20, wherein the feedback comprises label weights associated with labels assigned to the first item of content, an identification of a contribution to an adjustment of a weight of a generative model used to generate the generated content, and/or a token (In Col. 56, lines 19-21, Aravamudan discloses adjusting the connections and weights between nodes in adjacent layers of the neural network to produce the desired values at the output nodes.)
Regarding claim 26, Aravamudan in view of LIU and SONG discloses the element of claim 20. In addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, wherein the estimated contribution of the first item of content to the generated model comprises an estimated contribution of vocabulary choice, sentence structure, grammar and punctuation, tone and voice, themes and topics, and/or rhetorical devices to the generated model (The computer system as defined in Claim 1, wherein the estimated contribution of the first item of content to the generative model output comprises an estimated contribution of vocabulary choice, sentence structure, grammar and punctuation, tone and voice, themes and topics, and/or rhetorical devices to the generative model output.)
Regarding claim 27, Aravamudan in view of LIU and SONG discloses element of claim 20. In addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, wherein the estimated contribution of the first item of content to the generated content comprises an estimated contribution of symbols, shapes, motifs, and/or iconography to the generated content (In Col. 41, lines 40–51, Aravamudan discloses any symbols usable as textual data.)
Regarding claim 28, Aravamudan in view of LIU and SONG discloses element of claim 20. In addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, wherein the estimated contribution of the first item of content to the generated content comprises an estimated contribution to one or more claims of the generated content (In Col. 8, lines 26–29, Aravamudan discloses that the machine learning module described herein may generate one or more generative machine learning models that are trained on one or more sets of training data)
Regarding claim 29, Aravamudan in view of LIU and SONG discloses element of claim 20. ?I addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, wherein transmitting the feedback generated based at least in part on the estimated contribution of the first item of content to the generated content to one or more networked destinations, further comprises transmitting feedback to a plurality of networked destinations based at least in part on estimated percentage contributions of a plurality of items of content to the generated content (In Col. 61, lines 8–14, Aravamudan discloses that a plurality of user feedback scores may be averaged to determine an accuracy score. In some embodiments, a cohort accuracy score may be determined for particular cohorts of persons. For example, user feedback for users belonging to a particular cohort of persons may be averaged together to determine the cohort accuracy score for that particular cohort of persons and used as described above)
Regarding claim 30, Aravamudan in view of LIU and SONG discloses element of claim 20. ?In addition, Aravamudan disclose:
The computer-implemented method as defined in Claim 20, the method further comprising transmitting an aggregated feedback for a first period of time to the one or more networked destinations (In Col. 10, lines 20–24, Aravamudan discloses that GAN may evaluate the authenticity of the generated content by comparing it to real first generative model output. For example, a discriminator may distinguish between genuine and generated content and provide feedback to the generator to improve the model performance.)
Claims 7-8, 16, 18-19 and 25 are rejected under 35 U.S.C. 103 as being unpatentable over Aravamudan in view of LIU, SONG and further in view of WO 2025170579 A1, 02/07/2024, CARBUNE VICTOR, hereinafter referred to as CARBUNE.
Regarding claim 7, Aravamudan in view of LIU and SONG disclose the elements of claim 1. Aravamudan in view of LIU and SONG does not appear to disclose:
The computer system as defined in Claim 1, wherein the first item of content comprises text data
However, CARBUNE disclose the limitation (In paragraph [0085], CARBUNE VICTOR discloses that a first candidate's output can include text data, image data, audio data, structure data, latent encoding data, multimodal data, and/or other data.)
Aravamudan in view of LIU, SONG and are CARBUNE analogous pieces of art because all the references concern configuring a generative machine learning model using a syntactic interface. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify CARBUNE, with a plurality of content items associated with a plurality of respective resources and a plurality of quality scores associated with the plurality of respective resources as taught by CARBUNE. The motivation for doing so would have been to improve the efficiency and improvements in the functioning of a computing system (See [0067] of CARBUNE.)
Regarding claim 8, Aravamudan in view of LIU and SONG disclose the elements of claim 1. Aravamudan in view of LIU and SONG does not appear to disclose:
The computer system as defined in Claim 1, wherein the estimated contribution of the first item of content to the generative model output comprises an estimated contribution of style
However, CARBUNE disclose the limitation (In paragraph [0085], CARBUNE VICTOR discloses that each of the plurality of candidate model -generated outputs 318 may be associated with content, style, sequence, and/or logic associated with a different content item from the training dataset 316.)
Aravamudan in view of LIU, SONG and are CARBUNE analogous pieces of art because all the references concern configuring a generative machine learning model using a syntactic interface. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify CARBUNE, with a plurality of content items associated with a plurality of respective resources and a plurality of quality scores associated with the plurality of respective resources as taught by CARBUNE. The motivation for doing so would have been to improve the efficiency and improvements in the functioning of a computing system (See [0067] of CARBUNE.)
Regarding claim 16, Aravamudan in view of LIU and SONG disclose the elements of claim 1. Aravamudan in view of LIU and SONG does not appear to disclose:
The computer system as defined in Claim 1, wherein estimating the contribution of the first item of content to the generative model output further comprises estimating a style contribution of the first item of content to the generative model output
However, CARBUNE VICTOR disclose the limitation (In paragraph [0085], CARBUNE VICTOR discloses that each of the plurality of candidate model -generated outputs 318 may be associated with content, style, sequence, and/or logic associated with a different content item from the training dataset 316.)
Aravamudan in view of LIU, SONG and are CARBUNE analogous pieces of art because all the references concern configuring a generative machine learning model using a syntactic interface. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify CARBUNE, with a plurality of content items associated with a plurality of respective resources and a plurality of quality scores associated with the plurality of respective resources as taught by CARBUNE. The motivation for doing so would have been to improve the efficiency and improvements in the functioning of a computing system (See [0067] of CARBUNE.)
Regarding claim 18, Aravamudan in view of LIU and SONG disclose the elements of claim 1. Aravamudan in view of LIU and SONG does not appear to disclose:
The computer system as defined in Claim 1, wherein the computer system is operable to: generate a prompt instructing the generative model to use only specified document chunks in providing a response to the generated prompt
receive a response to the generated prompt from the generative model
and determine similarities of the response to the generated prompt to the specified document chunks
However, CARBUNE disclose the limitation:
The computer system as defined in Claim 1, wherein the computer system is operable to: generate a prompt instructing the generative model to use only specified document chunks in providing a response to the generated prompt (In paragraph [0046], CARBUNE VICTOR disclose the systems and methods can then obtain a prompt. The systems and methods can process the prompt with a generative model to generate a model-generated output responsive to the prompt)
receive a response to the generated prompt from the generative model (In paragraph [0046], CARBUNE VICTOR discloses the systems and methods and can then obtain a prompt. The systems and methods can process the prompt with a generative model to generate a model-generated output responsive to the prompt.)
determine similarities of the response to the generated prompt to the specified document chunks (In paragraph [0046], CARBUNE VICTOR discloses the systems and methods that can determine a ground truth example from the training dataset based on the plurality of quality scores. The systems and methods may then evaluate a loss function that evaluates a difference between the model generated output and the ground truth example and adjust one or more parameters of the generative model based on the loss function.)
Aravamudan in view of LIU, SONG and are CARBUNE analogous pieces of art because all the references concern configuring a generative machine learning model using a syntactic interface. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify CARBUNE, with a plurality of content items associated with a plurality of respective resources and a plurality of quality scores associated with the plurality of respective resources as taught by CARBUNE. The motivation for doing so would have been to improve the efficiency and improvements in the functioning of a computing system (See [0067] of CARBUNE.)
Regarding claim 19, Aravamudan in view of LIU and SONG disclose the elements of claim 1. Aravamudan in view of LIU and SONG does not appear to disclose:
The computer system as defined in Claim 1, wherein the computer system is operable to adjust an attribution score for at least one content source based at least in part on user feedback with respect to generative model outputs generated using content from the at least one content source (In paragraph [0114], CARBUNE VICTOR discloses that the computing system can adjust one or more parameters of the generative model based on the loss function. Adjusting one or more parameters can be utilized to adjust the probability predictions to increase the first probability and/or reduce the second probability. The adjustment may prioritize prediction probabilities for content items with higher quality signals.)
Regarding claim 25, Aravamudan in view of LIU and SONG disclose the elements of claim 20. Aravamudan in view of LIU and SONG does not appear to disclose:
The computer-implemented method as defined in Claim 20, the method further comprising: generating a prompt instructing a generative model to use only specified document chunks in providing a response to a prompt
receiving a response to the prompt from the generative model
determining similarities of the response to the generated prompt to the specified document chunks
However, CARBUNE disclose the limitation:
The computer-implemented method as defined in Claim 20, the method further comprising: generating a prompt instructing a generative model to use only specified document chunks in providing a response to a prompt (In paragraph [0046], CARBUNE VICTOR disclose the systems and methods can then obtain a prompt. The systems and methods can process the prompt with a generative model to generate a model-generated output responsive to the prompt)
receiving a response to the prompt from the generative model (In paragraph [0046], CARBUNE VICTOR discloses the systems and methods and can then obtain a prompt. The systems and methods can process the prompt with a generative model to generate a model-generated output responsive to the prompt.)
determining similarities of the response to the generated prompt to the specified document chunks (In paragraph [0046], CARBUNE VICTOR discloses the systems and methods that can determine a ground truth example from the training dataset based on the plurality of quality scores. The systems and methods may then evaluate a loss function that evaluates a difference between the model generated output and the ground truth example and adjust one or more parameters of the generative model based on the loss function.)
Aravamudan in view of LIU, SONG and are CARBUNE analogous pieces of art because all the references concern configuring a generative machine learning model using a syntactic interface. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify CARBUNE, with a plurality of content items associated with a plurality of respective resources and a plurality of quality scores associated with the plurality of respective resources as taught by CARBUNE. The motivation for doing so would have been to improve the efficiency and improvements in the functioning of a computing system (See [0067] of CARBUNE.)
Response to Arguments
The applicant's arguments filed 07/28/2025 have been fully considered, but in part are not persuasive.
Pertaining to Rejection under 101
The rejection of 35 USC § 101 has been withdrawn.
Pertaining to Rejection under 103
Applicant’s arguments in regard to the examiner’s rejections under 35 USC 103 are moot in view of the new grounds of rejection
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to EVEL HONORE whose telephone number is (703)756-1179. The examiner can normally be reached Monday-Friday 8 a.m. -5:30 p.m.
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, Mariela D Reyes can be reached at (571) 270-1006. 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.
EVEL HONORE
Examiner
Art Unit 2142
/Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142