DETAILED ACTION
This action is responsive to the filing of 4/18/24. Claims 1-20 are pending and have been considered below.
Allowable Subject Matter
Claims 7-10, 15-18 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is an examiner's statement of reasons for allowance. The prior art of record fails to disclose using a helper LLM for a specific support topic for a specific cluster from the ranked order, in combination with other limitations recited within the claimed context. The claims present a combination of limitations that differ from the cited art, and there is no reasonable combination of references that would teach it.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4, 6, 11-12, 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gao (2025/0252254) in view of Roberts (2021/0133462.)
Claim 1, 19-20: Gao discloses a method comprising:
transforming an original corpus of support call transcriptions for support calls received at a telecommunication provider at least in part by prompting a large language model from a series of models to summarize each support call transcript in the original corpus of support call transcripts for the support calls received at the telecommunication provider to output a summary corpus of large language model generated summaries of support call transcriptions (par. 20, causing the one or more computerized LLMs to generate summaries of each of the text transcripts);
vectorizing, by a sentence embeddings model, the summary corpus of large language model generated summaries of support call transcriptions such that an original vector corpus is produced with a respective vector for each large language model generated summary in the summary corpus of large language model generated summaries (par. 47, based on the customer journey phase-labeled transcript 180, the call analysis engine 150 can track the path of the conversation and the amount of time that was spent on each topic and/or phase of the conversation in order to generate a summary of the customer's “journey” during the call 172. This information can be stored in the form of one or more data structures (e.g., files, database records, data vectors, etc.));
extracting, by referencing the original vector corpus, from the summary corpus of large language model generated summaries of support call transcriptions a ranked ordering of client support topics for clusters within the summary corpus of large language model generated summaries of support call transcriptions (par. 100, Further, the taxonomy generation module 308 can identify clusters of intent labels that are particularly large (e.g., indicating that those intent labels represent topics or contexts that are frequently exhibited during calls));
resolving, by the telecommunication provider, the client support topics in an actual order that is determined at least in part based on the ranked ordering of client support topics for clusters within the summary corpus of large language model generated summaries of support call transcriptions (par. 101, Further, this intent label identification process allows the system to determine a narrowly tailored set of intent labels for each type of call (e.g., 5-20 categories, in some example implementations), rather than using a large number of intent labels that may be difficult for a user to differentiate between them.)
However, Gao does not explicitly disclose:
improving accuracy of at least one earlier model in the series of models based on feedback received from a later model in the series of models.
Roberts discloses a similar method for machine learning model, including:
improving accuracy of at least one earlier model in the series of models based on feedback received from a later model in the series of models (par. 134, where external data obtained, the external data is used to update the machine learning model. For example, the monitoring system 102 may treat user responses and sensor data as feedback to train the machine learning model (e.g., in order to improve later accuracy.))
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to combine the teachings of Gao and Roberts so as to update and train the machine learning model Roberts (par. 134.)
Claim 2: Gao and Roberts disclose the method of claim 1, further comprising generating the original corpus of support call transcriptions by transcribing the support calls received at the telecommunication provider (Gao Fig. 3: 306, Audio Transcription Module 306.)
Claim 3: Gao and Roberts disclose the method of claim 2, wherein transcribing comprises at least one of: removing personally identifiable information; updating punctuation; labelling with at least one sentiment label; or performing proofreading (par. 98, the taxonomy generation module 308 performs a label generation process 312 based on the determined intent(s).)
Claim 4: Gao and Roberts disclose the method of claim 1, wherein:
the large language model has been fine-tuned on a domain relating to the support calls received at the telecommunication provider;
the large language model is specific to the domain relating to the support calls received at the telecommunication provider;
the large language model has been optimized for generating summaries (par. 96, The taxonomy generation module 308 performs intent summarization process 310 based on the text transcripts 304. For instance, the intent summarization process 310 can include determining the intent of one or more users during each of the calls (e.g., using the generative AI module 152));
or
prompting the large language model is performed using a prompt format for generating summaries that has been selected as superior from among multiple tested prompt formats for generating summaries.
Claim 6: Gao and Roberts disclose the method of claim 1, further comprising labeling the clusters with the client support topics by prompting, for each respective cluster, a same or different large language model to generate a respective label based on reading a sample of the large language model generated summaries of support call transcriptions within the respective cluster (par. 21, determining the contextual categories can include clustering the candidate contextual categories into a plurality of clusters, and selecting the contextual categories based on the clusters; par. 100, taxonomy generation module 308 performs an intent label clustering process 314 based on the generated intent labels and identifies a subset of the intent labels for use in annotating calls. As an example, the taxonomy generation module 308 can obtain each of the generated intent labels, and cluster similar intent labels together (e.g., labels representing a similar semantic meaning or context).)
Claim 11: Gao and Roberts disclose the method of claim 1, wherein extracting the ranked ordering of client support topics for clusters within the summary corpus of large language model generated summaries of support call transcriptions comprises ranking a first respective cluster higher based on a first return on investment for the first cluster being higher than a second return on investment for a second cluster in the clusters.
Claim 12: Gao and Roberts disclose the method of claim 1, wherein the client support topics comprise at least two of: phone activations or transfers; troubleshooting electronic subscriber identity module activations; customers seeking account access (par. 46, a customer journey phase label can indicate that a text segment 176 corresponds to a customer identification phase); general confusion (par. 46, issue identification phase); or language barriers.
Claim(s) 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gao (2025/0252254) in view of Roberts (2021/0133462) and in further view of Doretto (2023/0281970.)
Claim 5: Gao and Roberts disclose the method of claim 1. However, Gao does not explicitly disclose, further comprising performing domain adaptation on the large language model.
Doretto discloses a similar method for LLMs, including:
further comprising performing domain adaptation on the large language model (par. 57, Domain adaptation allows for taking a high-performing model trained on a large dataset and adapting it to work on new data representing some target domain. This enables the model to take advantage of learned low-level features acquired from the larger source dataset while tuning to the specific features of the target dataset, requiring less target data.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to combine the teachings of Gao and Roberts with Doretto so as to enable the model to take advantage of learned low-level features acquired from the larger source dataset while tuning to the specific features of the target dataset, requiring less target data (par. 57, Doretto.)
Claim(s) 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gao (2025/0252254) in view of Roberts (2021/0133462) and in further view of Limon (2024/0420062.)
Claim 13: Gao and Roberts disclose the method of claim 1. However, Gao does not explicitly disclose wherein the sentence embeddings model comprises all-MiniLM-L6-v2.
Limon discloses a similar method for LLMs, including: wherein the sentence embeddings model comprises all-MiniLM-L6-v2 (par. 29, The SentenceTransformer package uses the encoder function of the all-Mini-LM-L6-v2 model to generate a dense vector, i.e., an input sentence embedding, for each one of the input sentences.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to combine the teachings of Gao and Roberts with Limon so as to generate a dense vector, i.e., an input sentence embedding, for each one of the input sentences (par. 29, Limon.)
Claim(s) 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gao (2025/0252254) in view of Roberts (2021/0133462) and in further view of Sun (2023/0401121.)
Claim 14: Gao and Roberts disclose the method of claim 12. However, Gao does not explicitly disclose further comprising performing dimensionality reduction on the original vector corpus to generate a reduced vector corpus.
Sun discloses a similar method for text analysis, including:
performing dimensionality reduction on the original vector corpus to generate a reduced vector corpus (par. 86, In the technical solution disclosed by the present application, by processing the English logs into the form of word groups and phrases, the amount of elements in the corpus is greatly reduced, thus the bag-of-word model is simplified and the dimension of the log feature vector is reduced.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to combine the teachings of Gao and Roberts with Sun so as to reduce data amount in the vector corpus.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: Grillo (12,556,658) generating smart topics from transcripts.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ANDREY BELOUSOV whose telephone number is (571) 270-1695 and Andrew.belousov@uspto.gov email. The examiner can normally be reached Mon-Friday EST.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Adam Queler, can be reached at telephone number 571-272-4140. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center and the Private Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from Patent Center or Private PAIR. Status information for unpublished applications is available through Patent Center and Private PAIR for authorized users only. Should you have questions about access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
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) Form at https://www.uspto.gov/patents/uspto-automated- interview-request-air-form.
/Andrey Belousov/
Primary Examiner
Art Unit 2172
4/28/26
Prosecution Insights