DETAILED ACTION
This communication is in response to the Amendments and Arguments filed on March 12, 2026. Claims 1-20 are pending and have been examined. Hence, this action has been made FINAL.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
The reply filed on March 12, 2026 has been entered.
With respect to the applicant’s arguments to claim rejections under 35 U.S.C § 101, Applicant has amended each of the independent claims and asserts that “claim 1 explicitly recites an improvement in functioning of a computer or technical field by retraining an LLM to improve transcribing or summarizing audio content, thereby improving the functioning of the processing system”. The examiner agrees that these newly added limitations overcome the rejection under 35 U.S.C. 101.
With respect to the applicant’s arguments to claim rejections under 35 U.S.C § 102 and 103, the applicant’s arguments with respect to claims 1-20 have been considered but are moot in view of new ground(s) of rejection caused by the amendments.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
Claims 1-20 are rejected under 35 U.S.C. 112(a) as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, had possession of the claimed invention.
Independent claim 1 has been amended to recite:
“obtaining a group of large language models (LLMs)” (emphasis added).
Applicant identifies paragraphs [0025] of the instant application as allegedly supporting these amendments.
These specification sections, however, fail to provide an adequate written description of the “a group of large language models” or training thereof. The term “models” is not interchangeable with the term “large language models,” as these two terms have distinct and differing definitions in the art. Applicant failed to show adequate support in their instant specification for these amendments in direct contradiction to the requirements of MPEP 2163(II)(A) and 2163.04. Furthermore, the support for these limitations is not apparent. The closest specification to supporting these amendments is ¶ [0014], “the process/procedure for generating the transcripts 110 and/or the summaries 114 may conform with, or incorporate, one or more aspects of artificial intelligence, machine learning, natural language processing (NLP), or the like.” Paragraph [0012] of the specification supports generating summaries using large language models, but there is no language in the specification that supports that a plurality of large language models are retrieved, trained, or utilized to generate summaries. Thus, the amended limitations relating to the “a group of large language models” and their use in summarization, transcription, and training are not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventors, at the time the application was filed, had possession of the claimed invention.
Independent claims 17 and 19 are similarly rejected due to parallel claim language.
Likewise, the further recitation of the “first LLM from the group of LLMs” and its association with processing the first audio content to generate a transcript in claim 1 is not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventors, at the time the application was filed, had possession of the claimed invention. The closest specification to supporting these amendments is ¶ [0014], “the process/procedure for generating the transcripts 110 and/or the summaries 114 may conform with, or incorporate, one or more aspects of artificial intelligence, machine learning, natural language processing (NLP), or the like.” There is no language in the specification that supports generating a transcript using a large language model. Thus, the amended limitations relating to the “first LLM” and its use in generating transcripts are not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventors, at the time the application was filed, had possession of the claimed invention.
Claims 2-16, 18, and 20 are rejected due to their dependency upon claims 1, 17, and 19.
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 1-4, 6-12, 15, 17, and 19 are rejected under 35 U.S.C. 103 as obvious over "Improving Automatic Summarization for Browsing Longform Spoken Dialog" (Li et al.) in view of US Patent Publication 20240412723 A1 (Kim et al.) in view of US Patent Publication 20250078829 A1 (Nguyen et al.).
Claim 1
Regarding claim 1, Li et al. disclose a device, comprising:
a processing system including a processor (Li et al. pg. 4, Section 4, Paragraph 1, "We utilize multiple different transformers including BART-L for dialog summarization, PEGASUS for short abstractive summarization and T5 for grammar correction. These transformers were primarily chosen based on the pre-trained models that were available to the public on HuggingFace [75], which is important in ensuring that our research is reproducible and maintainable." Using off-the-shelf pre-trained models implies the use of a processor); and
a memory that stores executable instructions (Li et al. pg. 4, Section 4, Paragraph 1, "We utilize multiple different transformers including BART-L for dialog summarization, PEGASUS for short abstractive summarization and T5 for grammar correction. These transformers were primarily chosen based on the pre-trained models that were available to the public on HuggingFace [75], which is important in ensuring that our research is reproducible and maintainable." Using off-the-shelf pre-trained models implies the use of memory that stores instructions. See Appendix A.4.1 for pseudocode instructions) that, when executed by the processing system, facilitate performance of operations, the operations comprising:
obtaining an audio sample (Li et al. pg. 4, Section 3.3, Paragraph 1, "To properly evaluate our system, we prepared a dataset of 25 transcripts consisting of audio recordings to use as test data.")
generating text associated with the audio sample (Li et al. pg. 4, Section 3.3, Paragraph 1, "Transcripts were transcribed with speaker diarization using Google Speech-to-Text.");
obtaining a group of large language (LLMs) (Li et al. pg. 5, Section 4.2, Paragraph 1-2, "The baseline hierarchical summarization system is adapted from [37] ... We refer to the "baseline" hierarchical summarization instance as Baseline and an instance utilizing the framework containing our contributions as System. The Baseline and System both use two summarization language models a BART-L model that is finetuned on the SamSum Corpus [18] to handle larger segmented transcript chunks, and a PEGASUS paraphrase model for smaller inputs (30 words or less)."); …
obtaining first audio content associated with a first video (Li et al. pg. 4, Section 3.3, Paragraph 1, "To properly evaluate our system, we prepared a dataset of 25 transcripts consisting of audio recordings to use as test data.") [conference call];
processing the first audio content to generate a transcript utilizing speech-to-text technology (Li et al. pg. 4, Section 3.3, Paragraph 1, "Transcripts were transcribed with speaker diarization using Google Speech-to-Text.") [and a first LLM from the group of LLMs];
generating at least one summary (Li et al. pg. 1, Section 1, Paragraph 3, "First, the transcript is broken into 256- character chunks, then each chunk is summarized.") utilizing a second LLM from the group of LLMs based on the transcript (Li et al. pg. 5, Section 4.2, Paragraph 2, "We refer to the "baseline" hierarchical summarization instance as Baseline and an instance utilizing the framework containing our contributions as System. The Baseline and System both use two summarization language models a BART-L model that is finetuned on the SamSum Corpus [18] to handle larger segmented transcript chunks, and a PEGASUS paraphrase model for smaller inputs (30 words or less)." Any of the Baseline model or the System model are considered analogous to a second LLM);
identifying at least one error or inconsistency in the transcript or the at least one summary (Li et al. pg. 1, Section 1, Paragraph 6, "To improve readability we track entities and impute ambiguous coreferences to eliminate vague references. To improve accuracy we employ guided text decoding with contradiction assessment to enhance summary correctness." Imputing ambiguous coreferences or employing contradiction assessment is considered analogous to identifying at least one error or inconsistency in a summary) based on a third LLM from group of LLMs (Li et al. pg. 5, Section 4.2, Paragraph 2, "System performs Steps 1-4 (Fig 1, Section 4.3-4.5)." System is considered analogous to a third LLM); and
automatically implementing, based on the identifying, a correction or a clarification in respect to the at least one error or inconsistency in [both the transcript and] the at least one summary (Li et al. pg. 6, Section 4.4.1, Paragraph 1, "given a [summary]
s
i
and a context
L
c
o
n
t
e
x
t
= [
s
i
-
3
,
s
i
-
2
,
s
i
-
1
], any identifed coreference from
L
c
o
n
t
e
x
t
to
s
i
is imputed into
s
i
." See Table 2, which illustrates imputing a vague coreference with appropriate context) to generate [an adjusted transcript and] an adjusted at least one summary (Li et al. pg. 6, Table 2, Rows “Imputation” and “Grammer Fixed” both illustrate a generated adjusted at least one summary)….
Li et al. do not explicitly disclose all of detecting errors in a transcription.
However, Kim et al. disclose obtaining an audio sample (Kim et al. ¶ [0048], "The microphone 112 may receive spoken audio data from users 102 (as well as other sources, such as the display device 108). ");
generating text associated with the audio sample (Kim et al. ¶ [0057]-[0058], "when a user provides audio input, an ASR engine analyzes the audio input, recognizes the speech, and outputs a transcript, such as the text corresponding to the audio input. ... Transcription knowledge graph processing module 130 can utilize data from past sessions of the ASR engine to form a voice graph that can be analyzed to determine a correlation between a mis-transcription (error text) and the correct transcription (correct text).");
obtaining a [group of large language models (LLMs)] model (Kim et al. ¶ [0058], "the voice graph can be used to train machine learning (ML) embedded model algorithms to generate numerical representations of an entity.");
training the [group of [LLMs] model with the audio sample and the text (Kim et al. ¶ [0058], "the voice graph can be used to train machine learning (ML) embedded model algorithms to generate numerical representations of an entity.");
obtaining first audio content (Kim et al. ¶ [0061], "Block diagram 300 illustrates audio input 310 being received by ASR engine 320.") [associated with a first video conference call];
processing the first audio content to generate a transcript utilizing speech-to-text technology (Kim et al. ¶ [0061], "ASR engine 320 can generate ASR output 325 that includes a transcription that may be a mis-transcription (e.g., error text) or a correct transcription (e.g., correct text.)") [and a first LLM from the group of LLMs]; …
identifying at least one error or inconsistency in the transcript (Kim et al. ¶ [0061], "Voice graph ASR error correction module 340 can access user log database 325 and entity database(s) 350 to determine mined pairs where a mined pair includes an error text and the corresponding correct text (e.g., (error text, correct text).) Voice graph ASR error correction module 340 can receive ASR output 325 and utilize the mined pairs to correct any mis-transcriptions in ASR output 325") [or the at least one summary based on a third LLM from group of LLMs];
automatically implementing, based on the identifying, a correction or a clarification in respect to the at least one error or inconsistency in [both] the transcript (Kim et al. ¶ [0061], " Voice graph ASR error correction module 340 can receive ASR output 325 and utilize the mined pairs to correct any mis-transcriptions in ASR output 325") [and the at least one summary] to generate an adjusted transcript (Kim et al. ¶ [0064], "If a mis-transcription is determined, ASR error corrector 440 can replace the mis-transcription in text 345 with the correct transcription, and provide text 345 to NLU system 360.") [and an adjusted at least one summary]; and
retraining at least one of the first [LLM] model and the second [LLM] model with [the first audio content,] the transcript [or the at least one summary,] and the at least one error or inconsistency to improve transcribing [or summarizing] second audio content (Kim et al. ¶ [0085], "A training algorithm for phoneme-embedding generator 510 could be a seq-to-seq model where the input error text (e.g., “jurassic park world domination”) is converted into the output correct text (e.g., “jurassic park world dominion”). Once trained, the embedding layer is taken as the phoneme-embedding generator. Thus, as new entities are created and new mined pairs are developed, the ML embedding model can adapt accordingly to more accurately select the intended entity as the correct entity than if the ML embedding model did not have access to the mined pairs database.").
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Li et al.’s summarization system to incorporate Kim et al.’s transcription and model retraining.
The suggestion/motivation for doing so would have been that, “changing an [off-the-shelf] ASR engine in those cases is difficult. Even if training data and source codes were made available, new training data (e.g., new pairs of (human voice, transcript)) would be needed and the new training data is time-consuming to collect,” as noted by the Kim et al. disclosure in paragraph [0026].
Li et al. in view of Kim et al. do not explicitly disclose all of transcribing using an LLM.
However, Nguyen et al. disclose obtaining first audio content associated with a first video conference call (Nguyen et al. ¶ [0043], "the meeting host may be able to request that the meeting be recorded or that a transcript of the audio streams be prepared... the recording or transcription functionality may be off-loaded to a dedicated server (or servers), e.g., cloud recording servers, for recording the audio and video streams."); and
processing the first audio content to generate a transcript utilizing speech-to-text technology (Nguyen et al. ¶ [0064], "From the utterance, the ASR functionality 350 generates multiple hypotheses from a beam search of ASR decoding. In some examples, the ASR functionality 350 may also generate a corresponding score for each hypothesis.") and a first LLM from the group of LLMs (Nguyen et al. ¶ [0071], "If the highest score does not satisfy the threshold(s), a LLM 380 is employed.... The LLM 380 receives the hypotheses generated by the ASR functionality 350 and a request to provide a corrected transcription of the best hypothesis identified before rescoring." ¶ [0088], "After the LLM 380 has received the inputted hypotheses and constraints 306, it operates to generate and output a finalized transcription.")….
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Li et al. in view of Kim et al. to include Nguyen et al.’s LLM-based transcription because such a modification is the result of simple substitution of one known element for another producing a predictable result. More specifically, Kim et al.’s ML-based transcription and Nguyen et al.’s LLM-based transcription perform the same general and predictable function, the predictable function being transcribing audio content into a text format. Since each individual element and its function are shown in the prior art, albeit shown in separate references, the difference between the claimed subject matter and the prior art rests not on any individual element or function but in the very combination itself - that is in the substitution of Kim et al.’s ML-based transcription by replacing it with Nguyen et al.’s LLM-based transcription. Thus, the simple substitution of one known element for another producing a predictable result renders the claim obvious.
Claim 2
Regarding claim 2, the rejection of claim 1 is incorporated.
Li et al. further disclose wherein the processing of the first audio content comprises using automatic speech recognition (ASR) to generate the transcript (Li et al. pg. 4, Section 3.3, Paragraph 1, "Transcripts were transcribed with speaker diarization using Google Speech-to-Text.").
Claim 3
Regarding claim 3, the rejection of claim 1 is incorporated.
Nguyen et al. further disclose wherein the obtaining of the first audio content comprises extracting the first audio content from a recording associated with the first video conference call (Nguyen et al. ¶ [0043], "the meeting host may be able to request that the meeting be recorded or that a transcript of the audio streams be prepared... the recording or transcription functionality may be off-loaded to a dedicated server (or servers), e.g., cloud recording servers, for recording the audio and video streams.").
Claim 4
Regarding claim 4, the rejection of claim 1 is incorporated.
Li et al. further disclose wherein the at least one summary comprises a plurality of summaries (Li et al. pg. 7, Section 4.5.1, Paragraph 1, "the system generates multiple summaries based on given parameters").
Claim 6
Regarding claim 6, the rejection of claim 1 is incorporated.
Kim et al. further disclose wherein the at least one error or inconsistency is included in the transcript (Kim et al. ¶ [0061], "Voice graph ASR error correction module 340 can access user log database 325 and entity database(s) 350 to determine mined pairs where a mined pair includes an error text and the corresponding correct text (e.g., (error text, correct text).) Voice graph ASR error correction module 340 can receive ASR output 325 and utilize the mined pairs to correct any mis-transcriptions in ASR output 325").
Claim 7
Regarding claim 7, the rejection of claim 1 is incorporated.
Li et al. further disclose wherein the at least one error or inconsistency is included in the summary (Li et al. pg. 1, Section 1, Paragraph 6, "To improve readability we track entities and impute ambiguous coreferences to eliminate vague references. To improve accuracy we employ guided text decoding with contradiction assessment to enhance summary correctness." Imputing ambiguous coreferences or employing contradiction assessment is considered analogous to identifying at least one error or inconsistency in a summary).
Claim 8
Regarding claim 8, the rejection of claim 1 is incorporated.
Li et al. further disclose a group of LLMs (Li et al. pg. 5, Section 4.2, Paragraph 1-2, "The baseline hierarchical summarization system is adapted from [37] ... We refer to the "baseline" hierarchical summarization instance as Baseline and an instance utilizing the framework containing our contributions as System. The Baseline and System both use two summarization language models a BART-L model that is finetuned on the SamSum Corpus [18] to handle larger segmented transcript chunks, and a PEGASUS paraphrase model for smaller inputs (30 words or less).").
Kim et al. further disclose modifying, based on the implementing, the [group of LLMs] model, resulting in a modified [group of LLMs] model (Kim et al. ¶ [0085], "A training algorithm for phoneme-embedding generator 510 could be a seq-to-seq model where the input error text (e.g., “jurassic park world domination”) is converted into the output correct text (e.g., “jurassic park world dominion”). Once trained, the embedding layer is taken as the phoneme-embedding generator. Thus, as new entities are created and new mined pairs are developed, the ML embedding model can adapt accordingly to more accurately select the intended entity as the correct entity than if the ML embedding model did not have access to the mined pairs database.").
Claim 9
Regarding claim 9, the rejection of claim 8 is incorporated.
Li et al. further disclose obtaining a second least one summary based on the [second] transcript (Li et al. pg. 7, Section 4.5.1, Paragraph 1, "the system generates multiple summaries based on given parameters").
Kim et al. further disclose obtaining third audio content (Kim et al. ¶ [0034], "User(s) 102 may operate with the media system 104 to select and consume media content by, for example, providing audio commands to request media content.") [associated with a second video conference call];
processing the third audio content to generate a second transcript (Kim et al. ¶ [0050], "The transcription knowledge graph processing module 130 that receives the audio data may operate to process and analyze the received audio data to recognize the user 102's audio command.") utilizing the speech-to-text technology and a modified first LLM from the modified group of LLMs (Kim et al. ¶ [0029], "The transcription knowledge graph system can adapt to correct new ASR mis-transcriptions (e.g., new error texts) over time, and can train the ML embedding model to work in dynamic domains and accommodate new entities (e.g., new movies, audio books, authors) based on the adaptations."); … and
identifying a second at least one error or inconsistency in the second transcript (Kim et al. ¶ [0080], "Based on example 480, if ASR output 325 included “uk versus old miss”, ASR error corrector 440 can update the transcription to produce “UK verses ole miss” as the transcription of text 345. In another example, if ASR output 325 included “jurrassic park world domination,” ASR error corrector 440 can update the transcription to produce “jurrassic park world dominion” as the transcription of text 345.") or the second at least one summary based on a modified third LLM from the group of LLMs (Kim et al. ¶ [0029], "The transcription knowledge graph system can adapt to correct new ASR mis-transcriptions (e.g., new error texts) over time, and can train the ML embedding model to work in dynamic domains and accommodate new entities (e.g., new movies, audio books, authors) based on the adaptations." Modified third LLM is considered the same model as the modified first LLM.).
Nguyen et al. further disclose obtaining third audio content associated with a second video conference call (Nguyen et al. ¶ [0043], "the meeting host may be able to request that the meeting be recorded or that a transcript of the audio streams be prepared... the recording or transcription functionality may be off-loaded to a dedicated server (or servers), e.g., cloud recording servers, for recording the audio and video streams.").
Claim 10
Regarding claim 10, the rejection of claim 9 is incorporated.
Li et al. further disclose wherein the obtaining of the group of LLMs comprises generating the group of LLMs (Li et al. pg. 4, Section 4, Paragraph 1, "Our summarization framework is modular, not requiring any specific summarization model, and integrates external knowledge from language models trained on various natural language tasks with dialog heuristic constraints to construct a robust and unsupervised abstractive summarization system." Constructing an abstractive summarization system that contains multiple LLMs is considered analogous to generating a group of LLMs.).
Claim 11
Regarding claim 11, the rejection of claim 8 is incorporated.
Li et al. further disclose a group of LLMs (Li et al. pg. 5, Section 4.2, Paragraph 1-2, "The baseline hierarchical summarization system is adapted from [37] ... We refer to the "baseline" hierarchical summarization instance as Baseline and an instance utilizing the framework containing our contributions as System. The Baseline and System both use two summarization language models a BART-L model that is finetuned on the SamSum Corpus [18] to handle larger segmented transcript chunks, and a PEGASUS paraphrase model for smaller inputs (30 words or less).").
Kim et al. further disclose wherein the modifying of the [group of LLMs] model comprises retraining one of the [group of LLMs] model (Kim et al. ¶ [0085], "A training algorithm for phoneme-embedding generator 510 could be a seq-to-seq model where the input error text (e.g., “jurassic park world domination”) is converted into the output correct text (e.g., “jurassic park world dominion”).").
Claim 12
Regarding claim 12, the rejection of claim 1 is incorporated.
Li et al. further disclose wherein the obtaining of the first audio content comprises extracting the first audio content audio from at least one file (Li et al. pg. 4, Section 3.3, Paragraph 1, "To properly evaluate our system, we prepared a dataset of 25 transcripts consisting of audio recordings to use as test data. Transcripts were transcribed with speaker diarization using Google Speech-to-Text.").
Claim 15
Regarding claim 15, the rejection of claim 1 is incorporated.
Li et al. further disclose wherein the identifying of the at least one error or inconsistency in the transcript or the at least one summary is further based on data or information obtained from at least one source (Li et al. pg. 7, Section 4.5.1, Paragraph 2, "By using existing word knowledge systems [52], we can quickly and computationally tractably determine whether or not
H
i
’s hallucinated entities are proper generalizations of existing entities or truly inconsistent with
R
i
" Existing word knowledge systems are considered analogous to a source).
Claim 17
Regarding claim 17, Li et al. disclose a non-transitory machine-readable medium, comprising executable instructions (Li et al. pg. 4, Section 4, Paragraph 1, "We utilize multiple different transformers including BART-L for dialog summarization, PEGASUS for short abstractive summarization and T5 for grammar correction. These transformers were primarily chosen based on the pre-trained models that were available to the public on HuggingFace [75], which is important in ensuring that our research is reproducible and maintainable." Using off-the-shelf pre-trained models implies the use of a some non-transitory machine-readable medium that comprises executable instructions. See Appendix A.4.1. for pseudocode).
The remaining limitations of claim 17 are similar in scope to that of claim 1 and therefore are rejected for similar reasons as described above.
Claim 19
Regarding claim 19, the limitations of claim 19 are similar to that of claim 1 and therefore are rejected for similar reasons as described above.
Claim 5 is rejected under 35 U.S.C. 103 as obvious over Li et al. in view of Kim et al. in view of Nguyen et al. as applied to claim 4 above, and further in view of US Patent Publication 20240086461 A1 (Varakin).
Claim 5
Regarding claim 5, the rejection of claim 4 is incorporated. Li et al. in view of Kim et al. in view of Nguyen et al. disclose all the elements of the claimed invention as stated above.
Li et al. in view of Kim et al. in view of Nguyen et al. do not explicitly disclose all of targeting different audiences using multiple summaries.
However, Varakin discloses wherein a first summary of the plurality of summaries and a second summary of the plurality of summaries are different from one another based on the first summary being targeted to a first audience and the second summary being targeted to a second audience that is different from the first audience (Varakin ¶ [0016], "In a company setting, as another example, different departments may request/prefer different summaries tailor to each department's function. Accordingly, the system and methods described herein can generate a plurality of different candidate summaries for each audio interaction that account for these types of user preferences.").
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Li et al. in view of Kim et al. in view of Nguyen et al. to include Varakin’s audience-based summaries.
The suggestion/motivation for doing so would have been that, “ it would be beneficial to generate summaries that are tailored to individual user preferences, or preferences of a business having particular product lines or providing particular services. For example, a customer service agent may require different information in a summary for different types of products or related to different types of services,” as noted by the Varakin disclosure in paragraph [0016].
Claims 13-14, 16, 18, and 20 are rejected under 35 U.S.C. 103 as obvious over Li et al. in view of Kim et al. in view of Nguyen et al. as applied to claim 1 above, and further in view of US Patent Publication 20230386472 A1 (Li).
Claim 13
Regarding claim 13, the rejection of claim 1 is incorporated. Li et al. in view of Kim et al. in view of Nguyen et al. disclose all the elements of the claimed invention as stated above.
Li et al. in view of Kim et al. in view of Nguyen et al. do not explicitly disclose all of an error or inconsistency corresponding to a name misspelling.
However, Li discloses wherein the at least one error or inconsistency corresponds to a misspelling of a name (Li ¶ [0033], "The text transcription 300 includes several spelling errors. A first spelling error 308A misinterprets the name ‘Soham’ as the words ‘so him.’ A second spelling error 308B misspells the name ‘Christina’ as the word ‘Kristina.’ A third spelling error 308C misinterprets the name ‘Soham’ again as the words ‘so him.’").
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Li et al. in view of Kim et al. in view of Nguyen et al. to incorporate Li’s application to meeting data.
The suggestion/motivation for doing so would have been that, “A natural language processing (NLP) model of the computer conference application 102 that is configured to generate the text transcription is trained using training data that primarily includes words in the English language and includes a limited number of words from other languages, such as people's names that are common in different non-English languages. Due to this factor and other factors, the text transcription produced by the computer conference application 102 has various spelling errors,” as noted by the Li disclosure in paragraph [0020].
Claim 14
Regarding claim 14, the rejection of claim 13 is incorporated.
Li further discloses wherein the name corresponds to a person (Li ¶ [0033], "The text transcription 300 includes several spelling errors. A first spelling error 308A misinterprets the name ‘Soham’ as the words ‘so him.’ A second spelling error 308B misspells the name ‘Christina’ as the word ‘Kristina.’ A third spelling error 308C misinterprets the name ‘Soham’ again as the words ‘so him.’").
Claim 16
Regarding claim 16, the rejection of claim 15 is incorporated.
Li et al. in view of Kim et al. in view of Nguyen et al. do not explicitly disclose all of the source being one of the specified types.
However, Li discloses wherein the at least one source is based on: listings of employees, identifications of participants in a meeting, phonebooks, contact logs, emails, voicemails, text messages, or any combination thereof (Li ¶ [0064], "context data 218 indicates that the misspelled text ‘CRISTINA’ was included in an utterance spoken by ‘CHRISTINA,’ which was directed to the other participant—i.e., ‘KRISTINA.’ Accordingly, the transcription search machine 220 determines that the misspelled text actually corresponds to the spelling ‘KRISTINA’ and can make appropriate corrections as desired." ¶ [0019], "the first user 100, a second user 106, and a third user 108 are work colleagues that are virtually meeting to discuss a work matter.").
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Li et al. in view of Kim et al. in view of Nguyen et al. to incorporate Li’s application to meeting data.
The suggestion/motivation for doing so is similar to the suggestion/motivation described above with respect to claim 13.
Claim 18
Regarding claim 18, the rejection of claim 17 is incorporated.
Li et al. further discloses wherein the at least one error includes a plurality of errors (Li et al. pg. 1, Section 1, Paragraph 6, "To improve readability we track entities and impute ambiguous coreferences to eliminate vague references. To improve accuracy we employ guided text decoding with contradiction assessment to enhance summary correctness." Ambiguous coreferences and contradictions together are considered analogous to identifying a plurality of errors).
Li et al. in view of Kim et al. in view of Nguyen et al. do not explicitly disclose all of the plurality of errors corresponding the names of persons or entities.
However, Li discloses wherein the at least one error includes a plurality of errors, wherein a first error of the plurality of errors corresponds to a name of a person, and wherein a second error of the plurality of errors corresponds to a name of a business or entity (Li ¶ [0033], "The text transcription 300 includes several spelling errors. A first spelling error 308A misinterprets the name ‘Soham’ as the words ‘so him.’ A second spelling error 308B misspells the name ‘Christina’ as the word ‘Kristina.’ A third spelling error 308C misinterprets the name ‘Soham’ again as the words ‘so him.’" A person is considered analogous to an entity).
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Li et al. in view of Kim et al. in view of Nguyen et al. to incorporate Li’s application to meeting data.
The suggestion/motivation for doing so is similar to the suggestion/motivation described above with respect to claim 13.
Claim 20
Regarding claim 20, the rejection of claim 19 is incorporated. Li et al. disclose all the elements of the claimed invention as stated above.
Li et al. in view of Kim et al. in view of Nguyen et al. do not explicitly disclose all of correcting a plurality of names in a transcript.
However, Li discloses wherein the implementing of the at least one correction comprises modifying a plurality of names included in the transcript (Li ¶ [0033], "The text transcription 300 includes several spelling errors. A first spelling error 308A misinterprets the name ‘Soham’ as the words ‘so him.’ A second spelling error 308B misspells the name ‘Christina’ as the word ‘Kristina.’ A third spelling error 308C misinterprets the name ‘Soham’ again as the words ‘so him.’" ¶ [0063], "the text transcription with the specified spelling based at least on a context data 218 of the audio session. For example, context data includes names of speakers attributed to different text segments in the text transcription"), the modifying resulting in a modified transcript (Li ¶ [0062], "the transcription search machine 220 is configured replace the associated portion of the text transcription with the specified spelling to generate a corrected text transcription 252.")
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Li et al. in view of Kim et al. in view of Nguyen et al. to incorporate Li’s application to meeting data.
The suggestion/motivation for doing so is similar to the suggestion/motivation described above with respect to claim 13.
Reference Cited
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure.
US Patent Publication 20240194203 A1 to Koo et al. discloses automatically detecting ASR errors using statistical probabilities.
US Patent Publication 20230223026 A1 to Grichnik et al. discloses automatically modifying a user’s transcription based on weighting ASR probabilities using the user’s social media data.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JACOB B VOGT whose telephone number is (571)272-7028. The examiner can normally be reached Monday - Friday 9:30am - 7pm EST.
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, Paras D Shah can be reached at (571)270-1650. 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.
/JACOB B VOGT/ Examiner, Art Unit 2653
/Paras D Shah/ Supervisory Patent Examiner, Art Unit 2653
05/06/2026
Prosecution Insights