Systems and method for vocabulary management in a natural learning framework

Patent metadata

Field	Value
Publication number	US-10956683-B2
Application number	US-201916356815-A
Country	US
Kind code	B2
Filing date	Mar 18, 2019
Priority date	Mar 23, 2018
Publication date	Mar 23, 2021
Grant date	Mar 23, 2021

Abstract

An agent automation system implements a virtual agent that is capable of learning new words, or new meanings for known words, based on exchanges between the virtual agent and a user in order to customize the vocabulary of the virtual agent to the needs of the user or users. The agent automation framework has access to a corpus of previous exchanges between the virtual agent and the user, such as one or more chat logs. New words and/or new meanings for known words are identified within the corpus and new word vectors are generated for these new words and/or new meanings for known words and added to refine a word vector distribution model. The refined word vector distribution model is then utilized by the agent automation system to interact with the user.

First claim

What is claimed is:

1. An agent automation system, comprising:
a memory configured to store:
a natural language understanding (NLU) framework;
a word vector distribution model; and
a chat log; and

a processor configured to execute instructions to cause the agent automation system to perform actions comprising:
extracting a plurality of utterances from the chat log;
segmenting each of the plurality of extracted utterances into one or more words;
identifying a new word of the one or more words from the plurality of extracted utterances, wherein the new word does not have an associated word vector stored in the word vector distribution model;
generating a new word vector for the new word;
updating the word vector distribution model to include the new word vector;
receiving an utterance; and
generating an annotated utterance tree of the utterance, wherein the annotated utterance tree comprises at least one node that is associated with the new word vector of the word vector distribution model.

2. The agent automation system of claim 1 , wherein the new word vector is generated based on a context in which the new word was used in the plurality of extracted utterances.

3. The agent automation system of claim 2 , wherein the NLU framework comprises an ontology service and a structure service, wherein the ontology service and the structure service are configured to determine an intended meaning of the new word based on the context in which the new word was used.

4. The agent automation system of claim 2 , wherein the new word vector is generated based on a plurality of uses of the new word in the chat log.

5. The agent automation system of claim 1 , wherein the new word vector is generated based on input received from a user, wherein the received input comprises a definition of the new word.

6. The agent automation system of claim 1 , wherein the instructions cause the agent automation system to perform actions comprising:
identifying a new meaning of a word of the one or more words, wherein the new meaning does not have an associated word vector stored in the word vector distribution model;
generating a new word vector for the new meaning; and
updating the word vector distribution model to include the new word vector.

7. The agent automation system of claim 1 , wherein the word vector distribution model comprises at least one word vector for each known meaning for a plurality of known words.

8. The agent automation system of claim 1 , wherein the NLU framework comprises a prosody subsystem configured to segment each of the plurality of extracted utterances into the one or more words.

9. The agent automation system of claim 1 , wherein the NLU framework comprises a vocabulary subsystem, a structure subsystem, and a prosody subsystem that cooperate to generate the annotated utterance tree of the utterance.

10. The agent automation system of claim 9 , wherein the instructions cause the agent automation system to perform actions comprising identifying the new word in the utterance by determining that the new word does not have an associated word vector stored in the word vector distribution model.

11. The agent automation system of claim 9 , wherein the instructions cause the agent automation system to perform actions comprising generating a response to the utterance.

12. The agent automation system of claim 8 , wherein the prosody subsystem is configured to segment each of the plurality of extracted utterances into the one or more words based on written prosody cues, wherein the written prosody cues comprise a rhythm, an emphasis, or a focus of the plurality of extracted utterances.

13. A method, comprising:
extracting a plurality of utterances from a chat log stored in memory;
segmenting each of the plurality of extracted utterances into one or more words;
identifying a new usage of a word of the one or more words from the plurality of extracted utterances that does not match an associated word vector of a word vector distribution model stored in the memory;
generating a new word vector for the new usage;
updating the word vector distribution model to include the new word vector receiving an utterance; and
generating an annotated utterance tree of the utterance, wherein the annotated utterance tree comprises at least one node that is associated with the new word vector of the word vector distribution model.

14. The method of claim 13 , wherein the new word vector is generated based on a context of the new usage of the word in the plurality of extracted utterances.

15. The method of claim 14 , wherein the new word vector is generated based on a plurality instances of the new usage of the word in the chat log.

16. The method of claim 13 , wherein the new word vector is generated based on input received from a user, wherein the received input comprises a definition of the new meaning.

17. The method of claim 13 , wherein the word vector distribution model comprises at least one word vector for each known meaning for a plurality of known words.

18. The method of claim 13 , comprising identifying the new word usage in the utterance.

19. The method of claim 13 , comprising generating a response to the utterance.

20. A non-transitory, computer-readable medium storing instructions that, when executed by one or more processors of a computing system, cause the computing system to:
extract a plurality of utterances from a chat log stored in memory;
segment each of the plurality of extracted utterances into one or more words or phrases;
identify a new usage of a word or phrase of the one or more words or phrases from the plurality of extracted utterances that does not match an associated word vector of a word vector distribution model stored in the memory;
generate a new word vector for the new usage of the word or phrase;
update the word vector distribution model to include the new word vector receive an utterance; and
generate an annotated utterance tree of the utterance, wherein the annotated utterance tree comprises at least one node that is associated with the new word vector of the word vector distribution model.

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

• G06F40/211

  Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars · CPC title

• G06N5/025

  Extracting rules from data · CPC title

• G10L2015/223

  Execution procedure of a spoken command · CPC title

• G10L15/22

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title