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 .
DETAILED ACTION
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 06/29/26 has been entered.
This office action is in response to correspondence 06/29/26 regarding application 18/668,748, in which claims 1, 12, 18, 21, and 22 were amended. Claims 2 and 13 had been previously cancelled. Claims 1, 3-12, and 14-22 are pending in the application and have been considered.
Response to Arguments
The examiner agrees with Applicant on page 8 that the amendments to claims 1, 3, 4, 6, 10, 12, and 14-19 and new claims 21-22 do not add new matter.
On pages 8-9 Applicant argues that the amended claims differ from Sosea because the biased PageRank algorithm scores produced in Sosea are allegedly not an emotion prediction vector output from an emotion classification model. These arguments are moot based on the new grounds for rejection based in part on the newly cited reference to Dimitriadis et al. (US 20140163960), which contains an explicit emotion classification model that outputs a vector of prediction scores.
Applicant’s arguments on page 9 regarding amended claims 21 and 22 are also moot in view of the new ground for rejection based in part on Dimitriadis, which contains an explicit emotion classification model that outputs a vector of prediction scores.
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 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 of this title, 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, 3-5, 11, 12, 14-16, 18, and 20-22 are rejected under 35 U.S.C. 103 as being unpatentable over Sosea et al. (“Unsupervised Extractive Summarization of Emotion Triggers”. arXiv:2306.01444v1 [cs.CL] 2 Jun 2023) in view of Dimitriadis et al. (US 20140163960).
Consider claim 1, Sosea discloses a computer-implemented method (we carry out our experiments on an Nvidia A5000 GPU, Section 5, page 7: Experimental Setup) comprising:
receiving, by one or more processors and from an emotion model, an emotion vector for a text segment from a data object, the emotion vector comprising a plurality of emotion scores respectively corresponding to a plurality of characters indicating a plurality of emotions (compute an emotion-specific score for each word, which is combined with a sentence-similarity module to produce one sentence score per emotion, page 3, Section 2: Unsupervised Extractive Summarization; Let P be a Reddit post, Section 4, page 4, Task Formulation; See the example Reddit post in Figure 1, page 1; a separate biased PageRank is run for each emotion by computing vectors of emotions scores for each emotion for each word and sentence, Section 4, pages 5-6; the scores respectively corresponding to character sets making up the emotion labels, e.g. “a” “n” “g” “e” and “r” for “anger”, “j” “o” and “y” for joy, etc., Table 3, page 9556);
generating, by the one or more processors, a domain-specific relevancy prediction for the text segment based on a category-relevant subset of the plurality of emotion scores that correspond to a category-specific character of the plurality of characters associated with a domain-specific summarization category (e.g. Rjoy, Figure 3, page 5; an emotion specific relevance is computed for each sentence to the “Joy” domain using the word and Joy category biased PageRank runs; this is the relevance to a “Joy” domain extractive summary, pages 5-6; the “j” character in the “joy” label is category specific to “joy”; it does not appear in any of the other emotion labels, Table 3, page 9556);
identifying, by the one or more processors, the text segment as relevant from the data object based on a comparison between the domain-specific relevancy prediction and a relevancy threshold (we use this per-emotion score to rank the sentences in the post P, and for an emotion e, we only select the sentences to be part of the final summary with final scores above threshold t, Section 4, Constructing the Summaries, Page 6); and
initiating, by the one or more processors, a performance of a machine learning summarization operation based on the text segment (we leverage models trained on emotion detection datasets to build our emotion-specific summaries for a post P by concatenating sentences that express the same emotions, page 7, Emotion-specific baselines) by providing the relevant text as an input to a machine learning summarization model to receive a summary for the data object (of the n sentence is Reddit post P, the sentences are ranked by relevancy with biased PageRank and used to construct the summaries, pages 5-6, Section 4).
Sosea does not specifically mention receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model.
Dimitriadis discloses receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model (ASR produces textual stream, i.e. transcript data object, and lexical analysis unit which is a classifier provides emotional tags and confidence scores, which are received by emotion state detection unit, [0081]-[0082], Fig 4, Fig 5).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model in order facilitate quick and efficient review of phone-based transactions, as suggested by Dimitriadis ([0005]). Doing so would have led to predictable results of improving training and quality control purposes, as suggested by Dimitriadis ([0005]). The references cited are analogous art in the same field of text analysis.
Consider claim 12, Sosea discloses a system comprising one or more processors; and at least one memory storing processor-exectuable instructions that, when executed by the one or more processors (a computer with memory coupled to a processor is inherent for carrying out the described experiments on an Nvidia A5000 GPU, Section 5, page 7: Experimental Setup), cause the one or more processors to:
receive, from an emotion model, an emotion vector for an text segment from a data object, the emotion vector comprising a plurality of emotion scores respectively corresponding to a plurality of characters indicating a plurality of emotions (compute an emotion-specific score for each word, which is combined with a sentence-similarity module to produce one sentence score per emotion, page 3, Section 2: Unsupervised Extractive Summarization; Let P be a Reddit post, Section 4, page 4, Task Formulation; See the example Reddit post in Figure 1, page 1; a separate biased PageRank is run for each emotion by computing vectors of emotions scores for each emotion for each word and sentence, Section 4, pages 5-6; the scores respectively corresponding to character sets making up the emotion labels, e.g. “a” “n” “g” “e” and “r” for “anger”, “j” “o” and “y” for joy, etc., Table 3, page 9556);
generate a domain-specific relevancy prediction for the text segment based on a category-relevant subset of the plurality of emotion scores that correspond to a category-specific character of the plurality of characters associated with a domain-specific summarization category (e.g. Rjoy, Figure 3, page 5; an emotion specific relevance is computed for each sentence to the “Joy” domain using the word and Joy category biased PageRank runs; this is the relevance to a “Joy” domain extractive summary, pages 5-6; the “j” character in the “joy” label is category specific to “joy”; it does not appear in any of the other emotion labels, Table 3, page 9556);
identify the text segment as relevant from the data object based on a comparison between the domain-specific relevancy prediction and a relevancy threshold (we use this per-emotion score to rank the sentences in the post P, and for an emotion e, we only select the sentences to be part of the final summary with final scores above threshold t, Section 4, Constructing the Summaries, Page 6); and
initiate a performance of a machine learning summarization operation based on the text segment (we leverage models trained on emotion detection datasets to build our emotion-specific summaries for a post P by concatenating sentences that express the same emotions, page 7, Emotion-specific baselines) by providing the relevant text as an input to a machine learning summarization model to receive a summary for the data object (of the n sentence is Reddit post P, the sentences are ranked by relevancy with biased PageRank and used to construct the summaries, pages 5-6, Section 4).
Sosea does not specifically mention receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model.
Dimitriadis discloses receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model (ASR produces textual stream, i.e. transcript data object, and lexical analysis unit which is a classifier provides emotional tags and confidence scores, which are received by emotion state detection unit, [0081]-[0082], Fig 4, Fig 5).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model for reasons similar to those for claim 1.
Consider claim 18, Sosea discloses one or more non-transitory computer-readable storage media including instructions that, when executed by one or more processors (a computer with memory storing instructions and coupled to a processor is inherent for carrying out the described experiments on an Nvidia A5000 GPU, Section 5, page 7: Experimental Setup), cause the one or more processors to:
receive, from an emotion model, an emotion vector for an text segment from a data object, the emotion vector comprising a plurality of emotion scores respectively corresponding to a plurality of characters indicating a plurality of emotions (compute an emotion-specific score for each word, which is combined with a sentence-similarity module to produce one sentence score per emotion, page 3, Section 2: Unsupervised Extractive Summarization; Let P be a Reddit post, Section 4, page 4, Task Formulation; See the example Reddit post in Figure 1, page 1; a separate biased PageRank is run for each emotion by computing vectors of emotions scores for each emotion for each word and sentence, Section 4, pages 5-6; the scores respectively corresponding to character sets making up the emotion labels, e.g. “a” “n” “g” “e” and “r” for “anger”, “j” “o” and “y” for joy, etc., Table 3, page 9556);
generate a domain-specific relevancy prediction for the text segment based on a category-relevant subset of the plurality of emotion scores that correspond to a category-specific character of the plurality of characters associated with a domain-specific summarization category (e.g. Rjoy, Figure 3, page 5; an emotion specific relevance is computed for each sentence to the “Joy” domain using the word and Joy category biased PageRank runs; this is the relevance to a “Joy” domain extractive summary, pages 5-6; the “j” character in the “joy” label is category specific to “joy”; it does not appear in any of the other emotion labels, Table 3, page 9556);
identify the text segment as relevant from the data object based on a comparison between the domain-specific relevancy prediction and a relevancy threshold (we use this per-emotion score to rank the sentences in the post P, and for an emotion e, we only select the sentences to be part of the final summary with final scores above threshold t, Section 4, Constructing the Summaries, Page 6); and
initiate a performance of a machine learning summarization operation based on the text segment (we leverage models trained on emotion detection datasets to build our emotion-specific summaries for a post P by concatenating sentences that express the same emotions, page 7, Emotion-specific baselines) by providing the relevant text as an input to a machine learning summarization model to receive a summary for the data object (of the n sentence is Reddit post P, the sentences are ranked by relevancy with biased PageRank and used to construct the summaries, pages 5-6, Section 4).
Sosea does not specifically mention receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model.
Dimitriadis discloses receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model (ASR produces textual stream, i.e. transcript data object, and lexical analysis unit which is a classifier provides emotional tags and confidence scores, which are received by emotion state detection unit, [0081]-[0082], Fig 4, Fig 5).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by receiving, from an emotion classification model, an emotion prediction vector for a text segment from a transcript data object, the emotion prediction vector comprising a plurality of emotion prediction scores output by the emotion classification model for reasons similar to those for claim 1.
Consider claim 3, Sosea discloses: generating a sentiment for the data object based on a subset of the plurality of emotion scores that correspond to one or more texts from the data object (the subset of scores for each sentence of the post for each particular emotion, Constructing the Summaries, page 6); and assigning the sentiment to the summary (given a post P, we label the post with emotion e if we identify any sentence s as a summary for e, “Emotion Detection”, pages 7-8).
Sosea does not specifically mention utterances from the transcript data object, and concluding subset of the plurality of emotion prediction scores that correspond to one or more concluding utterances from the transcript data object.
Dimitriadis discloses utterances from the transcript data object, and a concluding subset of the plurality of emotion prediction scores that correspond to a concluding utterance from the transcript data object (e.g. the last words of the sentence “this is unacceptable I will…”, Fig. 4, [0062]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by including utterances from the transcript data object and a concluding subset of the plurality of emotion prediction scores that correspond to one or more concluding utterances from the transcript data object for reasons similar to those for claim 1.
Consider claim 4, Sosea discloses generating the sentiment comprises: generating a plurality of sentiment bucket scores based on the subset of the plurality of emotion scores (a separate biased PageRank for each emotion leverages a custom importance function for each word per emotion, which may be thought of as a “bucket” score as it represents the importance of the word based on other important vertices in the graph pointing to it, page 5, Emotion Decomposition), wherein: (i) a sentiment bucket score comprises an aggregation of a bucket subset of the subset of the plurality of emotion scores that correspond to a bucket-specific character of the plurality of characters (aggregation of scores for the individual words in the vertices, pages 5-6, Emotion Decomposition; the “j” character in the “joy” label is bucket specific to “joy”; it does not appear in any of the other emotion labels, Table 3, page 9556); and (ii) the plurality of sentiment bucket scores corresponds to a predefined sentiment option of a plurality of predefined sentiment options (we run the biased PageRank for each emotion, i.e. Anger, Disgust, Fear, etc. pages 5-6); and identifying the sentiment from the plurality of predefined sentiment options based on the plurality of sentiment bucket scores (given a post P, we label the post with emotion e if we identify any sentence s as a summary for e, “Emotion Detection”, pages 7-8).
Sosea does not specifically mention a transcript, and concluding subset of the plurality of emotion prediction scores; and a comparison between the plurality of sentiment bucket scores.
Dimitriadis discloses a transcript, and concluding subset of the plurality of emotion prediction scores (e.g. the last words of the sentence “this is unacceptable I will…”, Fig. 4, [0062]); and a comparison between the plurality of sentiment bucket scores (determining whether the emotion changes based on the confidence score comparison, [0064]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by including a concluding subset of the plurality of emotion prediction scores; and a comparison between the plurality of sentiment bucket scores for reasons similar to those for claim 1.
Consider claim 5, Sosea discloses: receiving the data object (a post P, page 6, Constructing the Summaries); identifying the text segment from the data object (sentences s1, s2…, sn, page 6, Constructing the Summaries); and providing the text segment as an input to the emotion model to receive the emotion vector (we score a sentence si as the average relevance of its words for emotion e, pages 5-6, Emotion Decomposition).
Sosea does not specifically mention a transcript utterance; and providing the text segment as an input to the emotion classification model to receive the emotion prediction vector.
Dimitriadis discloses a transcript utterance; and providing the text segment as an input to the emotion classification model to receive the emotion prediction vector (ASR produces textual stream, i.e. transcript data object, and lexical analysis unit which is a classifier provides emotional tags and confidence scores, which are received by emotion state detection unit, [0081]-[0082], Fig 4, Fig 5).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by including a transcript utterance and providing the text segment as an input to the emotion classification model to receive the emotion prediction vector for reasons similar to those for claim 1.
Consider claim 11, Sosea discloses the domain-specific relevancy prediction comprises an aggregation of the category-relevant subset of the plurality of emotion scores (e.g. Rjoy, Figure 3, page 5; an emotion specific relevance is computed for each sentence to the “Joy” domain using the word and Joy category biased PageRank runs; an aggregation of the word contributions, pages 5-6).
Sosea does not specifically mention emotion prediction scores.
Dimitriadis discloses emotion prediction scores (emotional tags and confidence scores, [0081]-[0082], Fig 4, Fig 5).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea to include emotion prediction scores for reasons similar to those for claim 1.
Consider claim 14, Sosea discloses the one or more processors are further caused to: generate a sentiment for the data object based on a subset of the plurality of emotion scores that correspond to one or more texts from the data object (the subset of scores for each sentence of the post for each particular emotion, Constructing the Summaries, page 6); and assigning the sentiment to the summary (given a post P, we label the post with emotion e if we identify any sentence s as a summary for e, “Emotion Detection”, pages 7-8).
Sosea does not specifically mention utterances from the transcript data object, and concluding subset of the plurality of emotion prediction scores that correspond to a concluding utterance from the transcript data object.
Dimitriadis discloses utterances from the transcript data object, and a concluding subset of the plurality of emotion prediction scores that correspond to a concluding utterance from the transcript data object (e.g. the last words of the sentence “this is unacceptable I will…”, Fig. 4, [0062]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by including utterances from the transcript data object and a concluding subset of the plurality of emotion prediction scores that correspond to one or more concluding utterances from the transcript data object for reasons similar to those for claim 1.
Consider claim 15, Sosea discloses generating the sentiment comprises: generating a plurality of sentiment bucket scores based on the subset of the plurality of emotion scores (a separate biased PageRank for each emotion leverages a custom importance function for each word per emotion, which may be thought of as a “bucket” score as it represents the importance of the word based on other important vertices in the graph pointing to it, page 5, Emotion Decomposition), wherein: (i) a sentiment bucket score comprises an aggregation of a bucket subset of the subset of the plurality of emotion scores that correspond to a bucket-specific character of the plurality of characters (aggregation of scores for the individual words in the vertices, pages 5-6, Emotion Decomposition; the “j” character in the “joy” label is bucket specific to “joy”; it does not appear in any of the other emotion labels, Table 3, page 9556); and (ii) the plurality of sentiment bucket scores corresponds to a predefined sentiment option of a plurality of predefined sentiment options (we run the biased PageRank for each emotion, i.e. Anger, Disgust, Fear, etc. pages 5-6); and identifying the sentiment from the plurality of predefined sentiment options based on the plurality of sentiment bucket scores (given a post P, we label the post with emotion e if we identify any sentence s as a summary for e, “Emotion Detection”, pages 7-8).
Sosea does not specifically mention a transcript, and concluding subset of the plurality of emotion prediction scores; and a comparison between the plurality of sentiment bucket scores.
Dimitriadis discloses a transcript, and concluding subset of the plurality of emotion prediction scores (e.g. the last words of the sentence “this is unacceptable I will…”, Fig. 4, [0062]); and a comparison between the plurality of sentiment bucket scores (determining whether the emotion changes based on the confidence score comparison, [0064]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by including a concluding subset of the plurality of emotion prediction scores; and a comparison between the plurality of sentiment bucket scores for reasons similar to those for claim 1.
Consider claim 16, Sosea discloses the one or more processors are further caused to: receive the data object (a post P, page 6, Constructing the Summaries); identify the text segment from the data object (sentences s1, s2…, sn, page 6, Constructing the Summaries); and provide the text segment as an input to the emotion model to receive the emotion vector (we score a sentence si as the average relevance of its words for emotion e, pages 5-6, Emotion Decomposition).
Sosea does not specifically mention a transcript utterance; and providing the text segment as an input to the emotion classification model to receive the emotion prediction vector.
Dimitriadis discloses a transcript utterance; and providing the text segment as an input to the emotion classification model to receive the emotion prediction vector (ASR produces textual stream, i.e. transcript data object, and lexical analysis unit which is a classifier provides emotional tags and confidence scores, which are received by emotion state detection unit, [0081]-[0082], Fig 4, Fig 5).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea by including a transcript utterance and providing the text segment as an input to the emotion classification model to receive the emotion prediction vector for reasons similar to those for claim 1.
Consider claim 20, Sosea discloses the domain-specific relevancy prediction comprises an aggregation of the category-relevant subset of the plurality of emotion scores (e.g. Rjoy, Figure 3, page 5; an emotion specific relevance is computed for each sentence to the “Joy” domain using the word and Joy category biased PageRank runs; an aggregation of the word contributions, pages 5-6).
Sosea does not specifically mention emotion prediction scores.
Dimitriadis discloses emotion prediction scores (emotional tags and confidence scores, [0081]-[0082], Fig 4, Fig 5).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea to include emotion prediction scores for reasons similar to those for claim 1.
Consider claim 21, Sosea does not, but Dimitriadis discloses wherein an emotion prediction score, of the plurality of emotion prediction scores, is indexed to a corresponding character, of the plurality of characters, indicating a corresponding emotion, of the plurality of emotions, that is evaluated by the emotion classification model for the utterance text segment, wherein the emotion prediction score indicates a likelihood, determined by the emotion classification model, that the utterance text segment is indicative of the emotion indicated by the character to which the emotion prediction score is indexed (emotional tags and confidence scores indexed in table 400 for the recognized words, [0081]-[0082], Fig 4, Fig 5, indicating a likelihood e.g. that the emotion classifier predicted “angry” for “unacceptable” in the transcript).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea such that wherein an emotion prediction score, of the plurality of emotion prediction scores, is indexed to a corresponding character, of the plurality of characters, indicating a corresponding emotion, of the plurality of emotions, that is evaluated by the emotion classification model for the utterance text segment, wherein the emotion prediction score indicates a likelihood, determined by the emotion classification model, that the utterance text segment is indicative of the emotion indicated by the character to which the emotion prediction score is indexed for reasons similar to those for claim 1.
Consider claim 22, Sosea does not, but Dimitriadis discloses wherein an emotion prediction score, of the plurality of emotion prediction scores, is indexed to a corresponding character, of the plurality of characters, indicating a corresponding emotion, of the plurality of emotions, that is evaluated by the emotion classification model for the utterance text segment, wherein the emotion prediction score indicates a likelihood, determined by the emotion classification model, that the utterance text segment is indicative of the emotion indicated by the character to which the emotion prediction score is indexed (emotional tags and confidence scores indexed in table 400 for the recognized words, [0081]-[0082], Fig 4, Fig 5, indicating a likelihood e.g. that the emotion classifier predicted “angry” for “unacceptable” in the transcript).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea such that wherein an emotion prediction score, of the plurality of emotion prediction scores, is indexed to a corresponding character, of the plurality of characters, indicating a corresponding emotion, of the plurality of emotions, that is evaluated by the emotion classification model for the utterance text segment, wherein the emotion prediction score indicates a likelihood, determined by the emotion classification model, that the utterance text segment is indicative of the emotion indicated by the character to which the emotion prediction score is indexed for reasons similar to those for claim 1.
Claims 6, 8-10, 17, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Sosea et al. (“Unsupervised Extractive Summarization of Emotion Triggers”. arXiv:2306.01444v1 [cs.CL] 2 Jun 2023) in view of Dimitriadis et al. (US 20140163960), in further view of Churgin et al. (US 20240371367).
Consider claim 6, Sosea discloses: (i) the category-specific character is identified from a plurality of training texts based on the domain-specific summarization category (sentences which trigger the emotion e, for the training dataset, Section 3, page 3, the character “j” being specific to the “joy” emotion, Table 3, page 9556).
Sosea and Dimitriadis do not specifically mention a predictive correlation based on a similarity score between (a) a historical utterance of a training transcript that corresponds to the domain-specific summarization category and (b) a training summary of the training transcript.
Churgin discloses a predictive correlation based on a similarity score between (a) a historical utterance of a training transcript that corresponds to the domain-specific summarization category and (b) a training summary of the training transcript (correlation value between 0 and 1, between unlabeled transcript data and extractive summaries, [0057], [0067], [0091]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea and Dimitriadis by including discloses a predictive correlation based on a similarity score between (a) a historical utterance of a training transcript that corresponds to the domain-specific summarization category and (b) a training summary of the training transcript in order to increase summarization accuracy, as suggested by Churgin ([0002]), predictably improving customer interactions, as suggested by Churgin ([0003]). The references cited are analogous art in the same field of text analysis.
Consider claim 8, Sosea and Dimitriadis do not, but Churgin discloses the historical utterance is associated with a historical emotion prediction vector (using historical data pools to generate training data pools, including labels for speaker sentiment for each utterance, [0049-0050], [0121]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea and Dimitriadis such that the historical utterance is associated with a historical emotion prediction vector for reasons similar to those for claim 6.
Consider claim 9, Sosea and Dimitriadis do not, but Churgin discloses the domain-specific relevancy prediction includes a probability that the utterance text segment is associated with (i) an expression of intent, (ii) an expression of a resolution, or (iii) an expression of contextual information (each utterance is embedded into a dense vector fed to a multi-class classifier, which generates a distribution of confidence scores for the fifteen intent labels, [0033]; a “probability” is inherent in this type of classification).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea and Dimitriadis such that the domain-specific relevancy prediction includes a probability that the utterance text segment is associated with (i) an expression of intent, (ii) an expression of a resolution, or (iii) an expression of contextual information for reasons similar to those for claim 6.
Consider claim 10, Sosea and Dimitriadis do not, but Churgin discloses removing an utterance text segment from the transcript data object based on: (i) a location of the utterance text segment within the transcript data object or (ii) a content-based categorization of the utterance text segment (intent filters are applied to remove utterances unlikely to contribute to the substance of the call, [0036]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea and Dimitriadis by removing an utterance text segment from the transcript data object based on: (i) a location of the utterance text segment within the transcript data object or (ii) a content-based categorization of the utterance text segment for reasons similar to those for claim 6.
Consider claim 17, Sosea discloses: (i) the category-specific character is identified from a plurality of training texts based on the domain-specific summarization category (sentences which trigger the emotion e, for the training dataset, Section 3, page 3, the character “j” being specific to the “joy” emotion, Table 3, page 9556).
Sosea and Dimitriadis do not specifically mention a predictive correlation based on a similarity score between (a) a historical utterance of a training transcript that corresponds to the domain-specific summarization category and (b) a training summary of the training transcript.
Churgin discloses a predictive correlation based on a similarity score between (a) a historical utterance of a training transcript that corresponds to the domain-specific summarization category and (b) a training summary of the training transcript (correlation value between 0 and 1, between unlabeled transcript data and extractive summaries, [0057], [0067], [0091]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea and Dimitriadis by including discloses a predictive correlation based on a similarity score between (a) a historical utterance of a training transcript that corresponds to the domain-specific summarization category and (b) a training summary of the training transcript for reasons similar to those for claim 6.
Consider claim 19, Sosea and Dimitriadis do not, but Churgin discloses the instructions further cause the one or more processors to remove an utterance text segment from the transcript data object based on: (i) a location of the utterance text segment within the transcript data object or (ii) a content-based categorization of the utterance text segment (intent filters are applied to remove utterances unlikely to contribute to the substance of the call, [0036]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea and Dimitriadis by removing an utterance text segment from the transcript data object based on: (i) a location of the utterance text segment within the transcript data object or (ii) a content-based categorization of the utterance text segment for reasons similar to those for claim 6.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Sosea et al. (“Unsupervised Extractive Summarization of Emotion Triggers”. arXiv:2306.01444v1 [cs.CL] 2 Jun 2023) in view of Dimitriadis et al. (US 20140163960), in further view of Churgin et al. (US 20240371367), in further view of Ben-David et al. (US 20240346249).
Consider claim 7, Sosea discloses the training summary is generated (the training dataset, Section 3, page 3).
Sosea, Dimitriadis, and Churgin do not specifically mention using a large language model.
Ben-David discloses using a large language model (prompting the STLM, which may be GPTs, to generate a summary, [0021]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Sosea, Dimitriadis, and Churgin by using a large language model in order to improve efficiency and accuracy of automated summarization, as suggested by Ben-David ([0008]), with predictable improvements in business settings, as suggested by Ben-David ([0007]). The references cited are analogous art in the same field of text analysis.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Jesse Pullias whose telephone number is 571/270-5135. The examiner can normally be reached on M-F 8:00 AM - 4:30 PM. The examiner’s fax number is 571/270-6135.
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, Andrew Flanders can be reached on 571/272-7516.
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.
/Jesse S Pullias/
Primary Examiner, Art Unit 2655 09/30/26
Prosecution Insights