Robotic agent conversation escalation

Patent metadata

Field	Value
Publication number	US-10699708-B2
Application number	US-201815961393-A
Country	US
Kind code	B2
Filing date	Apr 24, 2018
Priority date	Apr 24, 2018
Publication date	Jun 30, 2020
Grant date	Jun 30, 2020

Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for a conversation monitoring system. In one aspect, a method includes receiving a conversation stream generated from a conversation between a user of a platform and a current robotic agent, where the current robotic agent is one of a plurality of robotic agents servicing the platform; processing the conversation stream to identify characteristics of the conversation; extracting a conversation knowledge graph that includes a plurality of first nodes that each correspond a concept discussed in the conversation; identifying a state change in the conversation from the identified characteristics of the conversation; merging the conversation knowledge graph with a domain knowledge graph into a merged knowledge graph; and replacing the current robotic agent with a replacement robotic agent from the robotic agents, the replacement robotic agent selected based on the percentage matching value of the merged knowledge graph.

First claim

What is claimed is:

1. A computer-implemented method for replacing a robotic agent that is engaged in a conversation with a platform user, the method being executed by one or more processors and comprising:
during a conversation between a user of a platform and a current robotic agent:
receiving a conversation stream generated from the conversation between the user of the platform and the current robotic agent, wherein the current robotic agent is one of a plurality of robotic agents servicing the platform, and wherein each of the robotic agents includes a plurality of features,
processing the conversation stream to identify characteristics of the conversation;
creating, based on the characteristics of the conversation, a conversation knowledge graph that includes a plurality of first nodes that each correspond to a concept discussed in the conversation,
identifying a state change in the conversation from the identified characteristics of the conversation,
based on the identification of the state change in the conversation, selecting a portion of the conversation knowledge graph by filtering one or more nodes of the plurality of first nodes from the conversation knowledge graph to provide a trimmed conversation knowledge graph, and
merging the trimmed conversation knowledge graph with a domain knowledge graph to provide a merged knowledge graph, the domain knowledge graph being distinct from the conversation knowledge graph and comprising a plurality of second nodes that represent features of a first robotic agent of the plurality of robotic agents and a plurality of third nodes that represent features of a second robotic agent of the plurality of robotic agents, and wherein the merged knowledge graph includes a first percentage matching value between the first nodes and the second nodes and a second percentage matching value between the first nodes and the third nodes; and

replacing the current robotic agent with a replacement robotic agent from the robotic agents to facilitate a conversation between the user and the replacement robotic agent, the replacement robotic agent selected as one of the first robotic agent and the second robotic agent based on the first percentage matching value and the second percentage matching value of the merged knowledge graph.

2. The method of claim 1 , wherein the characteristics of the conversation comprise a semantic sentiment of the conversation, a topic of the conversation, a tone of the conversation, a temporal dimension of the conversation, or a combination thereof.

3. The method of claim 1 , further comprising:
generating a confidence factor based on the percentage matching values, the confidence factor representing how closely the replacement robotic agent matches concepts discussed in the conversation.

4. The method of claim 1 , wherein the one or more nodes filtered from the conversation knowledge graph are determined to be irrelevant based on the state change.

5. The method of claim 4 , wherein the irrelevant first nodes are determined based on a temporal distance from the identified state change in the conversation.

6. The method of claim 4 , wherein the irrelevant first nodes are determined based on a relevance to the identified state change in the conversation.

7. The method of claim 1 , wherein the conversation knowledge graph includes a plurality of edges each connecting two of the first nodes, wherein each of the edges represent a relationship between the respective connected first nodes.

8. The method of claim 1 , wherein the identified state change in the conversation comprises a change in a tone of the conversation with the user.

9. The method of claim 1 , wherein the identified state change in the conversation comprises a change in a topic of conversation.

10. The method of claim 1 , wherein the conversation stream is generated from speech converted to text through natural language processing (NLP).

11. The method of claim 1 , wherein the robotic agents comprise chatbots.

12. The method of claim 1 , wherein the current robotic agent is replaced with a human agent instead of the replacement robotic agent, wherein the human agent is selected based on the percentage matching value of the merged knowledge graph, and wherein the domain knowledge graph includes a portion of the second nodes that each correspond to features describing each of the human agents.

13. One or more non-transitory computer-readable storage media coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
during a conversation between a user of a platform and a current robotic agent:
receiving a conversation stream generated from the conversation between the user of the platform and the current robotic agent, wherein the current robotic agent is one of a plurality of robotic agents servicing the platform and wherein each of the robotic agents includes a plurality of features,
processing the conversation stream to identify characteristics of the conversation;
creating, based on the characteristics of the conversation, a conversation knowledge graph that includes a plurality of first nodes that each correspond to a concept discussed in the conversation,
identifying a state change in the conversation from the identified characteristics of the conversation,
based on the identification of the state change in the conversation, selecting a portion of the conversation knowledge graph by filtering one or more nodes of the plurality of first nodes from the conversation knowledge graph to provide a trimmed conversation knowledge graph, and
merging the trimmed conversation knowledge graph with a domain knowledge graph to provide a merged knowledge graph, the domain knowledge graph being distinct from the conversation knowledge graph and comprising a plurality of second nodes that represent features of a first robotic agent of the plurality of robotic agents and a plurality of third nodes that represent features of a second robotic agent of the plurality of robotic agents, and wherein the merged knowledge graph includes a first percentage matching value between the first nodes and the second nodes and a second percentage matching value between the first nodes and the third nodes; and

replacing the current robotic agent with a replacement robotic agent from the robotic agents to facilitate a conversation between the user and the replacement robotic agent, the replacement robotic agent selected as one of the first robotic agent and the second robotic agent based on the first percentage matching value and the second percentage matching value of the merged knowledge graph.

14. The one or more non-transitory computer-readable media of claim 13 , wherein the operations further comprise:
generating a confidence factor based on the percentage matching values, the confidence factor representing how closely the replacement robotic agent matches concepts discussed in the conversation.

15. The one or more non-transitory computer-readable media of claim 13 , wherein the current robotic agent is replaced with a human agent instead of the replacement robotic agent, wherein the human agent is selected based on the percentage matching value of the merged knowledge graph, and wherein the domain knowledge graph includes a portion of the second nodes that each correspond to features describing each of the human agents.

16. The one or more non-transitory computer-readable media of claim 13 , wherein the identified state change in the conversation comprises a change i

Assignees

• Accenture Global Solutions Ltd

Inventors

• Lecue Freddy
• Gueret Christophe Dominique Marie

Classifications

• G10L15/22 Primary

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

• G10L25/63

  for estimating an emotional state · CPC title

• G06F16/9024 Primary

  Graphs; Linked lists (G06F16/9027 takes precedence) · CPC title

• G10L2015/223

  Execution procedure of a spoken command · CPC title

• G10L15/1815

  Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning · CPC title