Method and apparatus for generating dialogue state

Patent metadata

Field	Value
Publication number	US-12229519-B2
Application number	US-202217806086-A
Country	US
Kind code	B2
Filing date	Jun 9, 2022
Priority date	Jun 30, 2021
Publication date	Feb 18, 2025
Grant date	Feb 18, 2025

Abstract

A method for generating a dialogue state includes: acquiring a target dialogue state of a previous round of dialogue and dialogue information of a current round of dialogue; generating an initial dialogue state of the current round of dialogue according to the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue; and generating a target dialogue state of the current round of dialogue according to the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue.

First claim

What is claimed is:

1. A method for generating a dialogue state, performed by a terminal, comprising:
acquiring a target dialogue state of a previous round of dialogue and dialogue information of a current round of dialogue;
generating an initial dialogue state of the current round of dialogue according to the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue; wherein generating the initial dialogue state of the current round of dialogue according to the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue, comprises:
acquiring a first dialogue state generation model; and
obtaining the initial dialogue state of the current round of dialogue by inputting the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue into the first dialogue state generation model;
generating a target dialogue state of the current round of dialogue according to the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue; wherein generating the target dialogue state of the current round of dialogue according to the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue, comprises:
acquiring a second dialogue state generation model; and
obtaining the target dialogue state of the current round of dialogue by inputting the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue into the second dialogue state generation model;
wherein the first dialogue state generation model and the second dialogue state generation model are Transformer encoder-decoder pre-training models; and the first dialogue state generation model and the second dialogue state generation model comprise L transformer blocks respectively;
wherein the dialogue information of the current round of dialogue and the target dialogue state of the previous round of dialogue are inputted into the first dialogue state generation model, and bidirectional encoding and decoding is performed by L transformer blocks to generate the initial dialogue state of the current round of dialogue, and the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue are inputted into the second dialogue state generation model, and bidirectional encoding and decoding is performed by the L transformer blocks to generate the target dialogue state of the current round of dialogue.

2. The method of claim 1 , wherein model parameters of the first dialogue state generation model are the same as model parameters of the second dialogue state generation model.

3. An electronic device, comprising:
at least one processor; and
a memory communicatively connected to the at least one processor; wherein,
the memory is stored with instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is configured to:
acquire a target dialogue state of a previous round of dialogue and dialogue information of a current round of dialogue;
generate an initial dialogue state of the current round of dialogue according to the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue; and
generate a target dialogue state of the current round of dialogue according to the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue;
wherein the at least one processor is further configured to:
acquire a first dialogue state generation model; and obtain the initial dialogue state of the current round of dialogue by inputting the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue into the first dialogue state generation model;
acquire a second dialogue state generation model; and obtain the target dialogue state of the current round of dialogue by inputting the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue into the second dialogue state generation model;
wherein the first dialogue state generation model and the second dialogue state generation model are Transformer encoder-decoder pre-training models; and the first dialogue state generation model and the second dialogue state generation model are Transformer encoder-decoder pre-training models; and the first dialogue state generation model and the second dialogue state generation model comprise L transformer blocks respectively;
wherein the dialogue information of the current round of dialogue and the target dialogue state of the previous round of dialogue are inputted into the first dialogue state generation model, and bidirectional encoding and decoding is performed by L transformer blocks to generate the initial dialogue state of the current round of dialogue, and the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue are inputted into the second dialogue state generation model, and bidirectional encoding and decoding is performed by the L transformer blocks to generate the target dialogue state of the current round of dialogue.

4. The electronic device of claim 3 , wherein model parameters of the first dialogue state generation model are the same as model parameters of the second dialogue state generation model.

5. A non-transitory computer-readable storage medium stored with computer instructions, wherein, the computer instructions are configured to enable a computer to perform a method for generating a dialogue state, the method comprising:
acquiring a target dialogue state of a previous round of dialogue and dialogue information of a current round of dialogue;
generating an initial dialogue state of the current round of dialogue according to the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue; wherein generating the initial dialogue state of the current round of dialogue according to the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue, comprises:
acquiring a first dialogue state generation model; and
obtaining the initial dialogue state of the current round of dialogue by inputting the target dialogue state of the previous round of dialogue and the dialogue information of the current round of dialogue into the first dialogue state generation model;
generating a target dialogue state of the current round of dialogue according to the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue; wherein generating the target dialogue state of the current round of dialogue according to the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue, comprises:
acquiring a second dialogue state generation model; and
obtaining the target dialogue state of the current round of dialogue by inputting the initial dialogue state of the current round of dialogue and the dialogue information of the current round of dialogue into the second dialogue state generation model;
wherein the first dialogue state generation model and the second dialogue state generation model are Transformer encoder-decoder pre-training models; and the first dialogue state generation model and the second dialogue state generation model comprise L transformer blocks respectively;
wherein the dialogue information of

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Tian Xin
• Huang Liankai
• Lin Yingzhan
• Bao Siqi
• He Huang
• Wang Fan
• Sun Shuqi
• Huang Shiwei

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/08

  Learning methods · CPC title

• G06F40/289

  Phrasal analysis, e.g. finite state techniques or chunking · CPC title

• G06F40/35 Primary

  Discourse or dialogue representation · CPC title

• G06F16/3329 Primary

  Natural language query formulation · CPC title