An Assistant Decision-Making System Design And Implementation Of Intention Analysis Based On China Mobile Communication Corporation Data

With the development of the mobile Internet,there are a large number of interactive data generated by various industries every day.Through the use of these data,intelligent Assistant Decision-Making System(ADSS)for the convenience of people's lives have emerged,such as Baidu map intelligent navigation system and Siri,etc.However,in the telecom industry,the function of intelligent assistant decision-making is still in its infancy.In this context,China Mobile Communication Corporation(CMCC)expects to analyze the dialogue intention of users to rationally develop business,improve the business system and build smart voice customer service system to better serve customers.Therefore,I designed and implemented a system to assist CMCC in business decision making through intention analysis of dialogue data.The core task of this system is intention analysis,which is divided into two parts.One is call-reason classification task based on people-to-people dialogue.To solve this problem,this article proposes a hierarchical neural network model.It adds a variety of word representations and uses ensemble scheme.The other part is the intent recognition of human-to-machine dialogue.It includes two sub-tasks,the intention classification and slot filling.For the intention classification,this article uses multiple simple classification models to obtain features of different dimensions.It also uses model ensemble scheme to obtain final results.For the slot filling,according to the semantic slot setting by CMCC,the large-scale corpus is summed up and the rules are extracted.On the one hand,the representation method of dialogue and sentence follows the characteristics of data features.The performance of these models is expected to be as good as possible.On the other hand,the real-time processing of the system is considered,so that the models should be simple.The results on test data show that these two targeted modeling methods solve the proposed problem quickly and effectively.Compared with the original model used by CMCC,the accuracy of call-reason classification task has increase by 7.26%and the intention recognition has increase by 5.19%.At the same time,through the analysis of the requirements of the system,the overall implementation architecture of the system is designed.The main functional modules of the assistant decision-making system are based on the MVC pattern.By integrating core system function modules and system architecture,CMCC can finally make a reasonable business.The quality of intelligent voice customer service has been improved.It makes the customer satisfaction improved 6.59%in sample survey in a short period.