Research And Implementation Of Informal Language Generation In Dialogue System

Natural Language Generation is one of the vital researching areas of the Natural Language Processing. The purpose of Natural Language Generation is to generate text in natural language automatically. It plays a very important role in the applications of Natural Language Processing task, like Human Computer Interaction, Machine Translation, Automatice Summarrzation, etc. In this paper, we implement natural language generation technology in the application of human computer interaction dialogue system, in which the sentences are generated automaticlly by computer.Using the mainstream natural language generation methods could only generate formal style sentences or written language sentences upon most occasions. However in the real applications such as chatting robot, automatic question answering system of E-commerce company and so on, people use informal language(network language and oral language) for the most of the time. Hence, the research and implement for informal language generation in dialogue system appears particularly important and urgent.In this paper, we first propose and implement a method based on Chinese FrameNet (CFN) to generate natural language sentences, in which we generate chinese sentences using the extensible various frameworks in CFN, along with the input of the system.Secondly, we also bring the idea of offline construction of formal and informal parallel word dictionary using Bootstrapping method into reality. Meanwhile we present a novel method to calculate the Pinyin similarity, which measures the distance between the formal word and the corresponding informal word candidates more precisely and by consequences, improve the quality of offline construction of formal word and informal parallel word dictionary.In the practical application, we use the method and technology mentioned above to implement an informal language generation system. The system would take users’language style into account and generate formal or informal sentence feedback based on the situations through formal-informal parellel word dictionary. We implement the system into a scenario in which the user talk to the NAO robot and teach it some knowledge through conversation like it is a child. The system shows a good performance with lively words and clearly expressed.