Research On Task-Oriented Dialogue System In Instrument Field

With the rapid development of artificial intelligence,the application range of dialogue system is becoming wider and wider.As an important branch of human-computer dialogue system,the demand of task-based dialogue system is increasing day by day.In the field of instrument and meter,with the continuous development of instrument technology,the more complex the operation steps of related tasked,the higher the learning costs of users.Therefore,the application of task dialogue system in the field of instrumentation not only can help users to complete tasks efficiently,but also can promote the development of the industry.In view of the lack of personalization in task-based dialog system,this paper proposes a personalized task-based dialog system model based on task-tree.Using deep learning,natural language processing and other related technologies,The prototype system of the model is implemented on the basis of the open source framework Rasa.Firstly,the paper analyzes the phenomenon of lack of individuation in task-based dialogue system.To solve this phenomenon,a personalized task-based dialogue scheme,which is based on task-tree,is proposed.Different task tree structures formed by different tasks.According to the task tree structure,the corresponding personalized task-based dialogue scheme is proposed.The existing task-based dialogue data sets have the phenomenon of low data quality.Aiming at the problems such as lack of personalized data and incomplete data sets,a method of establishing personalized task-based dialogue data sets is proposed.In this paper,a personalized task-based dialog system model based on task-tree is designed.And to implement the model,this paper uses the open source framework Rasa.Secondly,the intention recognition and named entity recognition are realized by using BERT,Bi LSTM and CRF model.The task tree is constructed according to task-related knowledge,which provides theoretical basis for the establishment of dialogue data set and dialogue management.Meanwhile,a dialogue data set containing personalized data and a speech template are established in this paper,which are used for training the personalized task-based dialog system model.In the dialogue management module,the status information of task tree is added to dialogue state tracking module;this paper analyzes the process that executes a personalized conversation policy based on the task-tree and conversation state tracking information,and discusses the corresponding process of personalized dialogue strategy from different task-tree structures in detail.Finally,the personalized task-based dialogue system based on task-tree is applied to the field of instrument and meter.And based on the signal generator hardware experiment and Lab VIEW software,this paper achieves the prototype system.The prototype system can interact with the user through UI interface.Additional,it can help users to complete the tasks of generating frequency modulation signals with the signal generator and realizing the signal generator with Lab VIEW software,which verifies the feasibility of the model.