Font Size: a A A

Design And Implementation Of Knowledge-based Human-Machine Dialogue System

Posted on:2020-03-08Degree:MasterType:Thesis
Country:ChinaCandidate:G Y GuFull Text:PDF
GTID:2428330575957039Subject:Computer technology
Abstract/Summary:PDF Full Text Request
In recent years,human-machine dialogue system,as a basic application of artificial intelligence,has received more and more attention from academic and industrial.The task-oriented human-machine dialogue system can complete the basic repetitive work and save labor costs;the non-task human-machine dialogue system can provide a more intelligent and more anthropomorphic dialogue experience.Sequence-to-sequence model-based dialogue systems suffer from tending to generate commonplace and meaningless responses,and it is difficult to provide practical help to users.Introducing external knowledge can enrich the amount of information,improve the diversity of responses and alleviate high-frequency universal response problems.In this paper,we design and implement a knowledge-based end-to-end dialogue model.This paper extends the standard attention-based sequence-to-sequence model with knowledge-based attention,knowledge attention mechanism and candidate target entities prediction to make the external structured knowledge information can guide the generation of responses in the model decoding process.In addition,by introducing a hierarchical dialogue history encoder,we improve the model's ability to encode the real and complex dialog data and the quality of the model's response generated in complex dialogue scenarios.Finally,the effectiveness of the proposed model is verified by the simulation dialogue data and the open dialogue dataset in the film field.Experiments show that the proposed model can generate responses with more knowledge and richer information.This paper designs and implements a knowledge-based dialogue data construction system and a dynamic dialogue service system.The knowledge-based dialogue data construction system can provide the necessary data foundation for the training of the model by collecting the dialogue corpus related to the existing knowledge data,so that the model proposed in this paper can be extended to different fields.The dynamic dialogue service system provides the ability to customize the system structure and quickly deploy common human-machine dialog models through a flexible modular structure and a unified interface between modules.Finally,based on the previous work,this paper implements a human-machine dialogue system in the field of movie,which can reply the user's response containing rich movie-related knowledge according to the dialogue history during the dialogue process.
Keywords/Search Tags:dialogue system, seq2seq, attention mechanism, knowledge encoder
PDF Full Text Request
Related items