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
This office action is in response to application 18/753,242, which was filed 06/25/24. On 09/09/24, in a preliminary amendment, Applicant amended claims 1, 8, and 15, cancelled claims 2-3, 9-10, and 16-17, and added new claims 21-26. Claims 1, 4-8, 11-15, and 18-26 are pending in the application and have been considered.
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, 5, 7, 8, 12, 14, 15, 19, 22, 24, and 26 are rejected under 35 U.S.C. 103 as being unpatentable over Larcheveque et al. (US 20110238410) in view of He et al. (“Deep Semantic Role Labeling: What Works and What’s Next”. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 473–483 Vancouver, Canada, July 30- August 4, 2017).
Consider claim 1, Larcheveque discloses a computer-implemented method for automatically generating fine-grained call reasons from customer service call transcripts (automatically identifying top issues of customer calls, [0353], from transcripts of calls for customer support, [0018], including fine grained reasons such as needing to change a password, [0056], as functions implemented on a computer, [0027]), the computer-implemented method comprising:
extracting, by an event extraction system using a natural language processing (NLP) technique, a set of events from a set of text strings of speaker turns (e.g. concepts “how” and “change password”, which are a set of events since they were uttered, are extracted from the transcribed utterance “How does one change one’s password”, “How to change password”, and “Would you be so kind as to tell me how to modify my password?”, i.e. speaker turns, [0056] by semantic parsing, [0057], an NLP technique), wherein said extracting comprises:
feeding, by the event extraction system, the set of text strings of speaker turns to a semantic role labeling parser to assign a respective semantic role result to each speaker turn in the set of text strings of speaker turns (text from the transcribed utterances is parsed to generate semantic graphs, the nodes representing concepts and the edges semantic roles, [0018], [0019], [0041]); and
extracting, by the event extraction system, the set of events based on the respective semantic role result assigned to each speaker turn in the set of text strings of speaker turns (graph intersection metric assigns a proximity to a pair of graphs g1 and g2 by finding the most informative graph that subsumes both a subgraph of g1 and a subgraph of g2, [0143]; this is considered to “extract” the concept nodes common to the subgraphs based on the edges, i.e. semantic roles, since the found graph contains those nodes, [0165-0159]);
identifying, by a cluster generation system, a set of clusters of events based on the set of events (clustering the semantic graphs using a quality threshold based on the graph intersection proximity metric, [0218], [0246], [0250]; this is considered to cluster based on the common nodes, i.e. “set of events”);
labeling, by the cluster generation system, each cluster of events in the set of clusters of events to generate a set of labeled clusters of events (semantic clusters from the corpus of labels are assigned precise topic names in the form of intents in the decision tree, [0316]); and
assigning, by a cluster assignment system, each event in the set of events to a respective labeled cluster of events in the set of labeled clusters of events (the semantic graphs, and intersecting nodes are clustered using semantic clustering algorithm, which assigns semantic graphs to a cluster, [0218-0232], [0246], [0250]).
Larcheveque does not specifically mention a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network.
He discloses a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network (highway LSTM with four layers, Figure 1, page 474, trained for SRL, page 476, Section 3.2).
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 Larcheveque by utilizing a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network as in He in order to better predict long-distance dependencies, as suggested by He (page 473). Doing so would have led to predictable results of improved semantic parsing, as suggested by He (Section 6, page 481). The references cited are analogous art in the same field of natural language processing.
Consider claim 8, Larcheveque discloses non-transitory computer readable medium including instructions for causing a processor to perform operations (RAM storing program code that causes a processor to perform the functions, [0027]) for automatically generating fine-grained call reasons from customer service call transcripts (automatically identifying top issues of customer calls, [0353], from transcripts of calls for customer support, [0018], including fine grained reasons such as needing to change a password, [0056]), the operations comprising:
extracting, using a natural language processing (NLP) technique, a set of events from a set of text strings of speaker turns (e.g. concepts “how” and “change password”, which are a set of events since they were uttered, are extracted from the transcribed utterance “How does one change one’s password”, “How to change password”, and “Would you be so kind as to tell me how to modify my password?”, i.e. speaker turns, [0056] by semantic parsing, [0057], an NLP technique), wherein said extracting comprises:
feeding the set of text strings of speaker turns to a semantic role labeling parser to assign a respective semantic role result to each speaker turn in the set of text strings of speaker turns (text from the transcribed utterances is parsed to generate semantic graphs, the nodes representing concepts and the edges semantic roles, [0018], [0019], [0041]); and
extracting the set of events based on the respective semantic role result assigned to each speaker turn in the set of text strings of speaker turns (graph intersection metric assigns a proximity to a pair of graphs g1 and g2 by finding the most informative graph that subsumes both a subgraph of g1 and a subgraph of g2, [0143]; this is considered to “extract” the concept nodes common to the subgraphs based on the edges, i.e. semantic roles, since the found graph contains those nodes, [0165-0159]);
identifying a set of clusters of events based on the set of events (clustering the semantic graphs using a quality threshold based on the graph intersection proximity metric, [0218], [0246], [0250]; this is considered to cluster based on the common nodes, i.e. “set of events”);
labeling each cluster of events in the set of clusters of events to generate a set of labeled clusters of events (semantic clusters from the corpus of labels are assigned precise topic names in the form of intents in the decision tree, [0316]); and
assigning each event in the set of events to a respective labeled cluster of events in the set of labeled clusters of events (the semantic graphs, and intersecting nodes are clustered using semantic clustering algorithm, which assigns semantic graphs to a cluster, [0218-0232], [0246], [0250]).
Larcheveque does not specifically mention a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network.
He discloses a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network (highway LSTM with four layers, Figure 1, page 474, trained for SRL, page 476, Section 3.2).
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 Larcheveque by utilizing a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network as in He for reasons similar to those for claim 1.
Consider claim 15, Larcheveque discloses a computing system for automatically generating fine-grained call reasons from customer service call transcripts (automatically identifying top issues of customer calls, [0353], from transcripts of calls for customer support, [0018], including fine grained reasons such as needing to change a password, [0056], as functions implemented on a computer, [0027])), the computing system comprising:
a storage unit configured to store instructions (RAM storing instructions, [0027]);
a control unit, coupled to the storage unit, configured to process the stored instructions (processor executes the instructions, [0027]) to:
extract, using a natural language processing (NLP) technique, a set of events from a set of text strings of speaker turns (e.g. concepts “how” and “change password”, which are a set of events since they were uttered, are extracted from the transcribed utterance “How does one change one’s password”, “How to change password”, and “Would you be so kind as to tell me how to modify my password?”, i.e. speaker turns, [0056] by semantic parsing, [0057], an NLP technique), wherein said extracting comprises:
feed the set of text strings of speaker turns to a semantic role labeling parser to assign a respective semantic role result to each speaker turn in the set of text strings of speaker turns (text from the transcribed utterances is parsed to generate semantic graphs, the nodes representing concepts and the edges semantic roles, [0018], [0019], [0041]); and
extract the set of events based on the respective semantic role result assigned to each speaker turn in the set of text strings of speaker turns (graph intersection metric assigns a proximity to a pair of graphs g1 and g2 by finding the most informative graph that subsumes both a subgraph of g1 and a subgraph of g2, [0143]; this is considered to “extract” the concept nodes common to the subgraphs based on the edges, i.e. semantic roles, since the found graph contains those nodes, [0165-0159]);
identify a set of clusters of events based on the set of events (clustering the semantic graphs using a quality threshold based on the graph intersection proximity metric, [0218], [0246], [0250]; this is considered to cluster based on the common nodes, i.e. “set of events”);
label each cluster of events in the set of clusters of events to generate a set of labeled clusters of events (semantic clusters from the corpus of labels are assigned precise topic names in the form of intents in the decision tree, [0316]); and
assign each event in the set of events to a respective labeled cluster of events in the set of labeled clusters of events (the semantic graphs, and intersecting nodes are clustered using semantic clustering algorithm, which assigns semantic graphs to a cluster, [0218-0232], [0246], [0250]).
Larcheveque does not specifically mention a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network.
He discloses a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network (highway LSTM with four layers, Figure 1, page 474, trained for SRL, page 476, Section 3.2).
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 Larcheveque by utilizing a semantic role labeling parser trained, using a deep highway long short-term memory (LSTM) neural network as in He for reasons similar to those for claim 1.
Consider claim 5, Larcheveque discloses labeling, by the cluster generation system, each cluster of events using a graph-based sentence compression algorithm that generalizes lexical variations in the events of each cluster of events (semantic representation of intent compresses and generalizes utterances such as “How should I go about changing my password” and “Need to change my password. How do I do that?” into the same, condensed semantic representation capturing intent, [0056]-[0057], and semantic clusters from the corpus of labels are assigned precise topic names in the form of intents in the decision tree, [0316]).
Consider claim 7, Larcheveque discloses each event in the set of events comprises a constituent phrase comprising a subject, a verb, and an object (e.g. “I”, “modify”, and “password”, Fig 7, [0056]-[0058]).
Consider claim 12, Larcheveque discloses labeling each cluster of events using a graph-based sentence compression algorithm that generalizes lexical variations in the events of each cluster of events (semantic representation of intent compresses and generalizes utterances such as “How should I go about changing my password” and “Need to change my password. How do I do that?” into the same, condensed semantic representation capturing intent, [0056]-[0057], and semantic clusters from the corpus of labels are assigned precise topic names in the form of intents in the decision tree, [0316]).
Consider claim 14, Larcheveque discloses each event in the set of events comprises a constituent phrase comprising a subject, a verb, and an object (e.g. “I”, “modify”, and “password”, Fig 7, [0056]-[0058]).
Consider claim 19, Larcheveque discloses processing the stored instructions to label each cluster of events using a graph-based sentence compression algorithm that generalizes lexical variations in the events of each cluster of events (semantic representation of intent compresses and generalizes utterances such as “How should I go about changing my password” and “Need to change my password. How do I do that?” into the same, condensed semantic representation capturing intent, [0056]-[0057], and semantic clusters from the corpus of labels are assigned precise topic names in the form of intents in the decision tree, [0316]).
Consider claim 22, Larcheveque discloses: extracting the set of text strings from the customer service call transcripts (text including transcripts of calls for customer support, [0018]).
Consider claim 24, Larcheveque discloses: extracting the set of text strings from the customer service call transcripts (text including transcripts of calls for customer support, [0018]).
Consider claim 26, Larcheveque discloses: extract the set of text strings from the customer service call transcripts (text including transcripts of calls for customer support, [0018]).
Claims 4, 11, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Larcheveque et al. (US 20110238410) in view of He et al. (“Deep Semantic Role Labeling: What Works and What’s Next”. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 473–483 Vancouver, Canada, July 30- August 4, 2017), in further view of Li et al. (US 20170337474).
Consider claim 4, Larcheveque and He do not, but Li discloses: feeding, by the cluster generation system, the set of events to a role factor network trained to generate a set of event embeddings that capture (i) distributional information associated with the set of events and (ii) interactions between arguments within the set of events (the network shown in Fig. 2, a role factor network as it embeds semantic roles, generates embeddings capturing semantic roles of arguments in the sentences, [0019], and distributional information of the words across sentences, [0013]); and identifying, by the cluster generation system, the set of clusters of events based on the set of event embeddings (clustering module clusters embeddings to represent the semantic roles of the embeddings in each cluster, [0016]).
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 Larcheveque and He by feeding, by the cluster generation system, the set of events to a role factor network trained to generate a set of event embeddings that capture (i) distributional information associated with the set of events and (ii) interactions between arguments within the set of events and identifying, by the cluster generation system, the set of clusters of events based on the set of event embeddings in order to overcome known difficulties of using artificial intelligence to understand natural language, as suggested by Li ([0001]). Doing so would have led to predictable results of improved parsing of sentences having differing forms, as suggested by Li ([0001]). The references cited are analogous art in the same field of natural language processing.
Consider claim 11, Larcheveque and He do not, but Li discloses: feeding the set of events to a role factor network trained to generate a set of event embeddings that capture (i) distributional information associated with the set of events and (ii) interactions between arguments within the set of events (the network shown in Fig. 2, a role factor network as it embeds semantic roles, generates embeddings capturing semantic roles of arguments in the sentences, [0019], and distributional information of the words across sentences, [0013]); and identifying the set of clusters of events based on the set of event embeddings (clustering module clusters embeddings to represent the semantic roles of the embeddings in each cluster, [0016]).
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 Larcheveque and He by feeding the set of events to a role factor network trained to generate a set of event embeddings that capture (i) distributional information associated with the set of events and (ii) interactions between arguments within the set of events and identifying the set of clusters of events based on the set of event embeddings for reasons similar to those for claim 4.
Consider claim 18, Larcheveque and He do not, but Li discloses: feed the set of events to a role factor network trained to generate a set of event embeddings that capture (i) distributional information associated with the set of events and (ii) interactions between arguments within the set of events (the network shown in Fig. 2, a role factor network as it embeds semantic roles, generates embeddings capturing semantic roles of arguments in the sentences, [0019], and distributional information of the words across sentences, [0013]); and identify the set of clusters of events based on the set of event embeddings (clustering module clusters embeddings to represent the semantic roles of the embeddings in each cluster, [0016]).
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 Larcheveque and He by feeding the set of events to a role factor network trained to generate a set of event embeddings that capture (i) distributional information associated with the set of events and (ii) interactions between arguments within the set of events and identifying the set of clusters of events based on the set of event embeddings for reasons similar to those for claim 4.
Claims 6, 13, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Larcheveque et al. (US 20110238410) in view of He et al. (“Deep Semantic Role Labeling: What Works and What’s Next”. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 473–483 Vancouver, Canada, July 30- August 4, 2017), in further view of Jayaraman (US 20200349199).
Consider claim 6, Larcheveque discloses storing, by the cluster generation system, the labeled clusters of events (storage of the named clusters is inherent to displaying them on a graphical user interface for a human reviewer to examine clusters, [0328]).
Larcheveque and He do not specifically mention a cluster database.
Jayaraman discloses a cluster database (clusters in the database, [0234]).
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 Larcheveque and He by storing the clusters in a cluster database in order to process large amounts of data, as suggested by Jayaraman ([0003]). The references cited are analogous art in the same field of natural language processing.
Consider claim 13, Larcheveque discloses storing the labeled clusters of events (storage of the named clusters is inherent to displaying them on a graphical user interface for a human reviewer to examine clusters, [0328]).
Larcheveque and He do not specifically mention a cluster database.
Jayaraman discloses a cluster database (clusters in the database, [0234]).
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 Larcheveque and He by storing the clusters in a cluster database for reasons similar to those for claim 6.
Consider claim 20, Larcheveque discloses storing the labeled clusters of events (storage of the named clusters is inherent to displaying them on a graphical user interface for a human reviewer to examine clusters, [0328]).
Larcheveque and He do not specifically mention a cluster database.
Jayaraman discloses a cluster database (clusters in the database, [0234]).
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 Larcheveque and He by storing the clusters in a cluster database for reasons similar to those for claim 6.
Claims 21, 23, and 25 are rejected under 35 U.S.C. 103 as being unpatentable over Larcheveque et al. (US 20110238410) in view of He et al. (“Deep Semantic Role Labeling: What Works and What’s Next”. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 473–483 Vancouver, Canada, July 30- August 4, 2017), in further view of Sinha et al. (US 10216724).
Consider claim 21, Larcheveque discloses: preprocessing, by a speaker turn preprocessing system, the set of text strings of speaker turns before extracting the set of events from the set of text strings of speaker turns (text from the transcribed utterances is parsed, [0018], [0019], [0041]).
Larcheveque and He do not specifically mention at least one of clean, autopunctuate, or resolve co-references in the set of text strings.
Sinha discloses at least one of clean, autopunctuate, or resolve co-references in a set of text strings (cleaning the text strings, Col 4 lines 52-63, noting the claim language only requires “at least one of”).
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 Larcheveque and He by preprocessing, by a speaker turn preprocessing system, the set of text strings of speaker turns by clean, autopunctuate, or resolve co-references in the set of text strings in order to overcome the difficulties in natural language processing of user generated content identified by Sinha (Col 1-2 lines 50-2). The references cited are analogous art in the same field of natural language processing.
Consider claim 23, Larcheveque discloses: preprocessing, by a speaker turn preprocessing system, the set of text strings of speaker turns before extracting the set of events from the set of text strings of speaker turns (text from the transcribed utterances is parsed, [0018], [0019], [0041]).
Larcheveque and He do not specifically mention at least one of clean, autopunctuate, or resolve co-references in the set of text strings.
Sinha discloses at least one of clean, autopunctuate, or resolve co-references in a set of text strings (cleaning the text strings, Col 4 lines 52-63).
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 Larcheveque and He by preprocessing, by a speaker turn preprocessing system, the set of text strings of speaker turns by clean, autopunctuate, or resolve co-references in the set of text strings for reasons similar to those for claim 21.
Consider claim 25, Larcheveque discloses: preprocessing, by a speaker turn preprocessing system, the set of text strings of speaker turns before extracting the set of events from the set of text strings of speaker turns (text from the transcribed utterances is parsed, [0018], [0019], [0041]).
Larcheveque and He do not specifically mention at least one of clean, autopunctuate, or resolve co-references in the set of text strings.
Sinha discloses at least one of clean, autopunctuate, or resolve co-references in a set of text strings (cleaning the text strings, Col 4 lines 52-63).
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 Larcheveque and He by preprocessing, by a speaker turn preprocessing system, the set of text strings of speaker turns by clean, autopunctuate, or resolve co-references in the set of text strings for reasons similar to those for claim 21.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 10134389 Hakkani-Tur discloses clustering user utterance intents with semantic parsing
US 20060080101 Chotimongkol discloses spoken language understanding by using semantic role labeling
US 20200387574 Min discloses linguistically rich cross-lingual text event embeddings
Xu et al. (“Conversational Semantic Role Labeling”. IEEE/ACM TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 29, 2021) discloses a conversational semantic role labeling task in which an argument can be the dialog participants, a phrase in the dialog history, or the current sentence
Masumura et al. (“Online Call Scene Segmentation of Contact Center Dialogues based on Role Aware Hierarchical LSTM-RNNs”. Proceedings, APSIPA Annual Summit and Conference 2018) discloses neural network based role-aware hierarchical long short-term memory recurrent neural network based call segmentation
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 02/18/26