The Study And Application Of Task-oriented Dialogue System Based On Multi-round Interaction

Task-oriented Dialogue System is one of the important tasks in natural language processing.It has a wide range of applications in intelligent customer service and personal assistants in daily life.Its task is to return the response generated by the system according to the user's input,and realize the user's request or goal through multiple rounds of interaction.To accomplish this task,classic neural network models are used to build a task-oriented dialogue system to achieve dialogue state tracking and system response generation.However,such methods still face some challenges,such as the problem that the dialogue system built by the neural network depends on a large amount of labeled data,and how to effectively introduce external knowledge in the dialogue system to adapt to complex scenarios.In order to meet the above challenges,this paper proposes a task-oriented dialogue model based on domain adaptation and the introduction of external knowledge.By using the domain adaptive method to alleviate the lack of annotated corpus in the task-based dialogue system,the domain migration of the task-based dialogue model is realized.By using a method of introducing external knowledge,the reasoning result of the knowledge graph is added to the dialogue system,so that the dialogue system is adapted to complex fields.The main work of this article includes:1.A task-oriented dialogue model based on domain adaptation.This model can transfer knowledge in source domains to a target domain with limited training samples.Specifically,we design a domain adaptive filter in the encoder-decoder model to reduce useless features in source domains and preserve common features.A domain adaptive amplifier is designed to enhance the target domain impact.We evaluate our method on both synthetic dialog and human-human dialog datasets and achieve state-of-the-art results.2.A task-oriented dialogue model that introduced external knowledge.The current method of introducing external knowledge into the dialogue system using memory networks and key-value retrieval mechanisms can enable the dialogue system to cope with some complex scenarios.However,these methods often affect the end-to-end training process of the dialogue model and have poor interpretability.The method in this paper adds entity relationship and knowledge graph reasoning to the decision-making part of the dialog model,so that the dialog model can use domain expertise in complex domains,thereby improving the performance of the model.3.A dialogue system for medical aided diagnosis.The dialogue model that introduced external knowledge can have better performance than the baseline model in complex dialogue scenarios such as medical treatment.In this paper,the proposed model is applied to real dialogue scenarios,and a multi-round dialogue system for medical aided diagnosis is constructed.The system can provide medical workers with auxiliary diagnosis functions in the form of natural language interaction,which is used to improve the accuracy of disease diagnosis and shorten the medical treatment process.The experimental results and analysis show that the domain-adaptive model proposed in this paper can alleviate the problem of the dialog system relying on a large amount of training data to a certain extent,and can guarantee the performance of the dialog model in the target domain;the models that introduce external knowledge are in two The test of the data set has better performance than the baseline model.Experimental results verify the effectiveness of this method.