Session message processing with generating responses based on node relationships within knowledge graphs

Patent metadata

Field	Value
Publication number	US-11488599-B2
Application number	US-201916979844-A
Country	US
Kind code	B2
Filing date	Apr 6, 2019
Priority date	Apr 24, 2018
Publication date	Nov 1, 2022
Grant date	Nov 1, 2022

Abstract

The present disclosure provides method and apparatus for processing a message. A statement sentence message and a message processing parameter associated with a user's session message are obtained. One or more first statement sentence nodes that are semantic-matched with the statement sentence message are determined in the knowledge map. One or more second statement sentence nodes corresponding to the message processing parameters are obtained from the knowledge map, based on the node relationship properties of the first statement sentence nodes. A response is generated based at least in part on statement sentences of the one or more second statement sentence nodes. The generated response is provided to the user.

First claim

The invention claimed is:

1. A method for processing a message, comprising:
obtaining a statement sentence message and a message processing parameter that are associated with a user's session message;
identifying one or more first statement sentence nodes that are semantic-matched with the statement sentence message in a knowledge graph and include at least one of a source webpage address and a timestamp, wherein the knowledge graph includes a plurality of statement sentence nodes, each first statement sentence node has a statement sentence and a node relationship property;
obtaining, from the knowledge graph, one or more second statement sentence nodes that are indicated by the node relationship property of the one or more first statement sentence nodes corresponding to the message processing parameter, wherein obtaining one or more second statement sentence nodes further comprises:
obtaining, from the knowledge graph, candidate statement sentence nodes that are indicated by the node relationship properties of the one or more first statement sentence nodes corresponding to the message processing parameter;
sorting candidate statement sentence nodes using the source webpage address and/or the timestamp of the obtained candidate statement sentence nodes; and
determining the one or more second statement sentence nodes from the sorted candidate statement sentence nodes;

generating a response based at least in part on statement sentences of the one or more second statement sentence nodes; and
providing the generated response to the user.

2. The method of claim 1 , wherein the node relationship property comprises at least one of an inter-node causation relationship property, a node topic category property, and an inter-node topic correlation property.

3. The method of claim 2 , wherein the node relationship property comprises an inter-node synonymous relationship property, and the method further comprises:
obtaining one or more third statement sentence nodes that have synonymous relationship with the one or more second statement sentence nodes,
wherein generating a response based at least in part on h statement sentences of the one or more second statement sentence nodes comprises:
generating a response based at least in part on the statement sentences of the one or more second statement sentence nodes and the statement sentences of the one or more third statement sentence nodes.

4. The method of claim 2 , wherein the message processing parameter indicates obtaining a statement sentence as a premise or conclusion of the statement sentence message, and
obtaining, from the knowledge graph, one or more second statement sentence nodes that are indicated by the node relationship property of the one or more first statement sentence nodes corresponding to the message processing parameter comprises:
obtaining, from the knowledge graph, one or more second statement sentence nodes as the premise or conclusion of the one or more first statement sentence nodes, based on the node topic category properties of the one or more first statement sentence nodes.

5. The method of claim 4 , wherein the node relationship property comprises a node standpoint property, the message processing parameter further indicates obtaining a statement sentence based on the standpoint, and the method further comprises:
determining a node standpoint parameter associated with the session message; and
obtaining one or more fourth statement sentence nodes that are matched with the determined node standpoint parameter from the one or more second statement sentence nodes,
wherein generating a response based at least in part on the statement sentences of the one or more second statement sentence nodes comprises:
generating a response based at least in part on the statement sentences of the one or more fourth statement sentence nodes.

6. The method of claim 5 , wherein the node standpoint parameter is determined based on at least one of the following messages:
the session message;
a context message of the session message;
a user's standpoint preference information; and
a set standpoint processing rule.

7. The method of claim 2 , wherein the message processing parameter indicates obtaining a statement sentence that belongs to a same topic category as that of the statement sentence message, and
obtaining, from the knowledge graph, one or more second statement sentence nodes that are indicated by the node relationship property of the one or more first statement sentence nodes corresponding to the message processing parameter comprises:
obtaining, from the knowledge graph, one or more statement sentence nodes that belong to a same topic category as that of the one or more first statement sentence nodes as the one or more second statement sentence nodes, based on the inter-node causation relationship properties of the one or more first statement sentence nodes.

8. The method of claim 2 , wherein the message processing parameter indicates obtaining a statement sentence that has topic correlation with the topic category of die statement sentence message, and
obtaining, from the knowledge graph, one or more second statement sentence nodes that are indicated by the node relationship property of the one or more first statement sentence nodes corresponding to the message processing parameter comprises:
obtaining, from the knowledge graph, one or more statement sentence nodes that have topic correlation with the topic categories of the one or more first statement sentence nodes as the one or more second statement sentence nodes, based on the inter-node topic correlation properties of the one or more first statement sentence nodes.

9. The method of claim 1 , wherein obtaining a statement sentence message and a message processing parameter that are associated with a user's session message comprises:
exacting the statement message from the session message; and
performing intention analysis on the session message to obtain the message processing parameter.

10. The method of claim 1 , wherein a statement sentence node-complete sentence mapping is also stored in the knowledge map, and the method further comprises;
obtaining complete sentences corresponding to the one or more second statement sentence nodes, based on the statement sentence node-complete sentence mapping,
wherein generating a response based at least in part on the statement sentences of the one or more second statement sentence nodes comprises
generating a response based at least in part on the statement sentences of the one or more second statement sentence nodes and/or the obtained corresponding complete sentences.

11. An apparatus for processing a message, comprising:
a message obtaining module, for obtaining a statement sentence message and a message processing parameter that are associated with a user's session message;
a semantic-matching module, for identifying one or more first statement sentence nodes that are semantic-matched with the statement sentence message in a knowledge graph and include at least one of a source webpage address and a timestamp, wherein the knowledge graph includes a plurality of statement sentence nodes, each statement sentence node has a statement sentence and a node relationship property;
a node determining module, for obtaining, from the knowledge graph, one or more second statement sentence nodes that are indicated by the node relationship properties of the one or more first statement sentence nodes corresponding to the message processing parameter, wherein obtaining one or more second statement sentence nodes further comprises:
obtaining, from the knowledge graph, candidate statement sentence nodes that are indicated by

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Chen Ling
• Shi Yu
• Chen Yining
• Zeng Nanshan
• Li Dong

Classifications

• G06F16/367

  Ontology · CPC title

• G10L15/04

  Segmentation; Word boundary detection · CPC title

• G10L15/26

  Speech to text systems (G10L15/08 takes precedence) · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06F16/332

  Query formulation · CPC title