End-to-end dialogue system and method based on dialogue state orientation and knowledge base retrieval

Abstract

The invention discloses an end-to-end dialogue system and method based on dialogue state guidance and knowledge base retrieval. The system comprises: a dialogue historical information encoding moduleused for encoding sentence-level semantic information by adopting a Transformer structure and encoding dialogue-level semantic information by adopting a GRU structure so as to output a dialogue-leveldistributed vector according to input dialogue historical information; a dialogue state tracking module used for extracting round-level information, extracting the inter-round information to obtain the probability distribution of the slot value on the current slot, and taking the maximum probability as the final slot value; a knowledge base retrieval module used for determining an accessed database and determining attribute entries in the accessed database; and a dialogue reply decoding module used for introducing a copy mechanism to decode so as to alleviate the OOV problem, and controlling the source of the copy by using a gating mechanism. The system has the advantages of accurate database positioning, accurate reply and logical reasoning capability.

Description

technical field [0001] The invention relates to the technical field of speech recognition, in particular to an end-to-end dialogue system and method based on dialogue state guidance and knowledge base retrieval. Background technique [0002] Related technologies, (1) an end-to-end task-based dialogue system, including: a preprocessing module for converting user input data into word vectors, and embedding position information for the word vectors; a named body extraction module for Identifies the nomenclature in the user input data; the compilation module is used to receive the data transmitted by the preprocessing module and the nomenclature extraction module, and output the probability distribution; the dialogue history encoder module is used to input the dialogue history into the The compilation module; the decoding output module is used to output natural language sentences as the system's reply to the user input. The invention can realize an end-to-end dialogue system, r...

AI Technical Summary

Problems solved by technology

[0003] However, the existing task-based end-to-end dialogue systems can directly realize the mapping from dialogue history to reply generation with only one codec model. Due to its simple structure, the semantic extraction ability is weak, resulting in meaningless replies

At the same time, due to the lack of a suitable knowledge base import method in the end-to-end system, the knowledge base entity cannot exist in the reply, and the "task" in the task-based dialogue system cannot be solved, which greatly reduces the practicability and reasoning ability of the dialogue system.

Although there are memory network components used in the knowledge retrieval of end-to-end systems, due to the lack of a reasonable retrieval guidance process, the system still has the problem of retrieval errors in the knowledge base

Images