Research on Affective Computing Based on Graph Sentiment Dictionary

Affective computing is an emerging field in recent years, it is proposed to allow computers understanding human emotions. At present, the most commonly used method of affective computing with textual signals is based on a sentiment lexicons. The polarity of emotion words can be divided into two types: positivity and negativity. The general method is to store common emotional words and their polarities in a list, taking them out of the list for calculation when needed. This approach can be used in basic affective computing , but it cannot solve the problems of polysemy. In view of this situation, we propose a new approach that is using a graphic dictionary to store sentiment words. The graph data structure is composed of two parts: vertices and edges, and the weights of several edges connecting the same vertex can be different from each other. The feature of the graph can solve the problem that one word may not have only one sentiment tendencies in different context. This paper constructs a graph sentiment dictionary based on the characteristics of the graph, then extracts the targets, features, sentiment words in the text as the vertices, and extracts the relationship between them as the edges to build a graph sentiment dictionary. By this way, the problem of polysemy and the lack of keywords can be solved. Compared with the traditional emotional dictionary, the effect of the graph sentiment dictionary will be better.