Computer-implemented method for generating a response sentence by using a weight value of node

Patent metadata

Field	Value
Publication number	US-2016293162-A1
Application number	US-201615076654-A
Country	US
Kind code	A1
Filing date	Mar 22, 2016
Priority date	Apr 2, 2015
Publication date	Oct 6, 2016
Grant date	—

Abstract

A semantic network storage unit stores multiple nodes necessary for performing a task of generating a response sentence to user's speech while associating the nodes with each other. A natural language processor acquires speech information indicating content of the user's speech and identifies a primary node corresponding to the speech information from multiple nodes. A conversation generation unit selects one secondary node from multiple secondary nodes associated with the identified primary node, based on weight values each related to one of the multiple secondary nodes and generates a response sentence corresponding to the selected one secondary node.

First claim

What is claimed is:

1 . A computer implemented method comprising:
acquiring speech information indicating content of user's speech by using a processor; identifying a primary node corresponding to the speech information from multiple nodes which are stored in a memory by using the processor, the multiple nodes being necessary for performing a task of generating a response sentence to the user's speech while associating the multiple nodes with each other; selecting one secondary node from multiple secondary nodes associated with the identified primary node, based on weight values each related to one of the multiple secondary nodes by using the processor; and generating a response sentence corresponding to the selected one secondary node by using the processor.

2 . The computer implemented method according to claim 1 , wherein
each of the weight values represents a probability that each of the multiple secondary nodes was selected by the user in the past.

3 . The computer implemented method according to claim 2 , wherein
a secondary node the probability of which is larger than a predetermined value is selected from the multiple secondary nodes.

4 . The computer implemented method according to claim 2 , wherein
in a case that a secondary node the probability of which is larger than a predetermined value is not present in the multiple secondary nodes, a response sentence for prompting the user to select any of the multiple secondary nodes is generated.

5 . The computer implemented method according to claim 1 , wherein
information indicating an answer of the user to the response sentence is acquired, and the weight values are updated depending on whether or not the user's answer is an answer to select one secondary node from the multiple secondary nodes.

6 . The computer implemented method according to claim 1 , wherein
a weight value is related to a combination of a first secondary node of multiple secondary nodes associated with a first primary node of multiple primary nodes and each of multiple secondary nodes associated with a second primary node of the multiple primary nodes, it is determined whether the first secondary node is identified, and in a case that the first secondary node is identified, a second secondary node is selected from the multiple secondary nodes associated with the second primary node, based on the weight values related to the respective combinations of the first secondary node and the multiple secondary nodes associated with the second primary node.

7 . A non-transitory medium having thereon a program for causing a processor to execute operations comprising:
acquiring speech information indicating content of user's speech; identifying a primary node corresponding to the speech information from multiple nodes which are stored in a memory, the multiple nodes being necessary for performing a task of generating a response sentence to the user's speech while associating the multiple nodes with each other; selecting one secondary node from multiple secondary nodes associated with the identified primary node, based on weight values each related to one of the multiple secondary nodes; and generating a response sentence corresponding to the selected one secondary node.

8 . An apparatus comprising:
a processor; and a memory having thereon a program, the program causing the processor to execute operations including: acquiring speech information indicating content of user's speech; identifying a primary node corresponding to the speech information from multiple nodes which are stored in a recording medium, the multiple nodes being necessary for performing a task of generating a response sentence to the user's speech while associating the multiple nodes with each other; selecting one secondary node from multiple secondary nodes associated with the identified primary node, based on weight values each related to one of the multiple secondary nodes; and generating a response sentence corresponding to the selected one secondary node.

Assignees

• Panasonic Ip Man Co Ltd

Inventors

• Takahashi Viviane
• Endo Mitsuru

Classifications

• G10L15/22 Primary

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

• G06F40/30

  Semantic analysis · CPC title

• G10L13/043

  Physics · mapped topic

• G10L2015/225

  Feedback of the input speech · CPC title

• G10L15/10 Primary

  using distance or distortion measures between unknown speech and reference templates · CPC title