Written-modality prosody subsystem in a natural language understanding (NLU) framework

Patent metadata

Field	Value
Publication number	US-11238232-B2
Application number	US-201916298764-A
Country	US
Kind code	B2
Filing date	Mar 11, 2019
Priority date	Mar 23, 2018
Publication date	Feb 1, 2022
Grant date	Feb 1, 2022

Abstract

Present embodiment include a prosody subsystem of a natural language understanding (NLU) framework that is designed to analyze collections of written messages for various prosodic cues to break down the collection into a suitable level of granularity (e.g., into episodes, sessions, segments, utterances, and/or intent segments) for consumption by other components of the NLU framework, enabling operation of the NLU framework. These prosodic cues may include, for example, source prosodic cues that are based on the author and the conversation channel associated with each message, temporal prosodic cues that are based on a respective time associated with each message, and/or written prosodic cues that are based on the content of each message. For example, to improve the domain specificity of the agent automation system, intent segments extracted by the prosody subsystem may be consumed by a training process for a ML-based structure subsystem of the NLU framework.

First claim

What is claimed is:

1. An agent automation system, comprising:
a memory configured to store a written conversation log and natural language understanding (NLU) framework including a prosody subsystem; and
a processor configured to execute instructions of the NLU framework to cause the agent automation system to perform actions comprising:
processing, via the prosody subsystem, the written conversation log based on prosodic cues to divide the written conversation log into conversation channel groups, to divide the conversation channel groups into sessions, to divide the sessions into conversation segments, to divide the conversation segments into utterances, and to divide the utterances into intent segments, wherein the prosodic cues comprise temporal prosodic cues and written prosodic cues.

2. The system of claim 1 , wherein, to process the conversation log, the processor is configured to execute the instructions of the NLU framework to cause the agent automation system to perform actions comprising:
dividing the written conversation log into the conversation channel groups, dividing the conversation channel groups into the sessions, dividing the sessions into the conversation segments, and dividing the conversation segments into the utterances based on metadata prosodic cues.

3. The system of claim 1 , wherein the temporal prosodic cues comprise a temporal prosody cue that is based on a respective time associated with each message of the conversation log.

4. The system of claim 3 , wherein the temporal prosody cue comprises a time gap between the respective times associated with each message of the conversation log.

5. The system of claim 1 , wherein, to divide the utterances into the intent segments, the processor is configured to execute the instructions of the NLU framework to cause the agent automation system to perform actions comprising:
analyzing the utterances for the written prosodic cues; and
dividing the utterances into the intent segments based on the written prosodic cues.

6. The system of claim 1 , wherein the written prosodic cues comprise punctuation, emojis, emphasis, or linguistic structure.

7. The system of claim 1 , wherein the written prosodic cues comprise an interrupt, a change in topic, a change in context, or a combination thereof.

8. The system of claim 1 , wherein the processor is configured to execute the instructions of the NLU framework to cause the agent automation system to perform actions comprising:
providing the intent segments as inputs to a training process for a machine-learning (ML)-based parser of the NLU framework, wherein, within the training process, the NLU framework is configured to apply a plurality of other parsers of the NLU framework to generate a plurality of utterance trees for each intent segment, and in response to determining that a majority of the plurality of utterance trees for a particular intent segment are the same utterance tree, update a model of the ML-based parser such that the ML-based parser generates the same utterance tree for the particular intent segment.

9. The system of claim 1 , wherein the processor is configured to execute the instructions of the NLU framework to cause the agent automation system to perform actions comprising:
providing the utterances as inputs to a training process for a vocabulary subsystem of the NLU framework, wherein, within the training process, the utterances are used to generate a plurality of word vectors of a refined word vector distribution model that replaces a word vector distribution model of the vocabulary subsystem, wherein the NLU framework is configured to use the refined word vector distribution model to determine a suitable word vector for words of received natural language requests.

10. The system of claim 1 , wherein the processor is configured to execute the instructions of the NLU framework to cause the agent automation system to perform actions comprising:
providing the intent segments as inputs to a semantic mining pipeline of the NLU framework, wherein the semantic mining pipeline is configured to:
generate intent vectors for the intent segments;
generate meaning clusters of intent vectors based on distances between the intent vectors;
detect stable ranges of cluster radius values for the meaning clusters; and
generate an intent/entity model from the meaning clusters and the stable ranges of cluster radius values, wherein the intent/entity model stores relationships between a representative intent of each of the meaning clusters and corresponding intent segments as sample utterances, and wherein the NLU framework is configured to use the intent/entity model to classify intents in received natural language requests.

11. The system of claim 1 , wherein the processor is configured to execute the instructions of the NLU framework to cause the agent automation system to perform actions comprising:
providing the sessions, the conversational segments, or a combination thereof, as inputs to a persona of a reasoning agent/behavior engine (RA/BE) of the NLU framework, wherein RA/BE is configured to generate an episode frame tree set in a persona context database of the persona based on each of the sessions, the conversational segments, or the combination thereof, wherein the episode frame tree set comprises an episode start time and an episode end time that are heuristically determined from the sessions, the conversational segments, or the combination thereof.

12. The system of claim 1 , wherein the processor is configured to execute the instructions of the NLU framework to cause the agent automation system to perform actions comprising:
receiving, from a persona of a RA/BE of the NLU framework, a new message that is part of a conversation between the persona and a user;
providing a first indication to the persona of the RA/BE in response to determining that the conversation is a continuation of a previous episode of conversation between the persona and the user; and
providing a second indication to the persona of the RA/BE in response to determining that the conversation is a new conversation episode.

13. A method of operating a prosody subsystem of a natural language understanding (NLU) framework, comprising:
dividing a conversation log comprising plurality of messages into a plurality of conversation channel groups based on a first set of prosodic cues;
dividing each of the plurality of conversation channel groups into a plurality of sessions based on a second set of prosodic cues;
dividing each of the plurality of sessions into a plurality of conversation segments based on a third set of prosodic cues;
dividing each of the plurality of conversation segments into a plurality of utterances based on a fourth set of prosodic cues;
dividing each of the plurality of utterances into a plurality of intent segments based on a fifth set of prosodic cues, wherein the second, third, fourth, and fifth sets of prosodic cues comprise temporal prosodic cues, written prosodic cues, or a combination thereof; and
providing the plurality of intent segments, the plurality of utterances, the plurality of conversation segments, or the plurality of sessions, or a combination thereof, as inputs to processes of the NLU framework.

14. The method of claim 13 , wherein the first set of prosodic cues comprises metadata prosodic cues that are based on a conversation channel associated with each of the plurality of messages of the conversation log, and wherein the second set of prosodic cues and third set of prosodic cues comprise the temporal cues that are based on a time associated with each of the plurality of messages of the co

Assignees

• Servicenow Inc

Inventors

• Sapugay Edwin
• Madamala Anil Kumar
• Naboka Maxim
• Sunkara Srinivas Satyasai
• Santos Lewis Savio Landry
• Subbarao Murali B

Classifications

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06N3/006

  based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO] · CPC title

• G06F40/205

  Parsing · CPC title

• G06N5/025

  Extracting rules from data · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title