The Research Of Sentiment Analysis Techniques For Short-Texts

With the advent of the information era, the continuous upgrading of information technology, Internet and mobile communication become the important styles of exchanging messages and expressing personal emotion in people's lives. Instant Message Software and mobile phone become the popular tools in the area of the information communication, following from them, there are large scale of message which imply the useful information and express emotion among people as a new text style. Because of their limited length, concise expression and rich emotion, we call them short-text which is widely used in mobile phone message, microblog text, BBS comments, and instant message records. Meanwhile, most short-texts are full of the user subjective expression on special topics, so the work on sentiment analysis in short-text has important practical value for us.Combing with the characteristics of short-texts ,sentiment analysis in short-text is a specific application of text classification, it belongs to the domain of natural language processing. This paper makes use of the findings of natural language processing to design and implement the system of sentiment analysis for short-texts. The paper's work contains:1) Proposing a model of sentiment analysis for short-texts. The paper analyzes the process of the natural language processing, combining with specific need of sentiment analysis for short-texts, the paper makes special analysis and description on text pre-processing, text feature extraction, semantic similarity and so on. Finally, the paper proposes the model of sentiment analysis for short-texts based on work above.2) Finishing the construction of the experimental system of sentiment analysis for short-texts. The paper researches the important models based on the model of sentiment analysis for short-texts. In the feature extraction model, the paper proposes to use the emotional words as text feature items to represent short-texts; in semantic similarity computing model, the paper considers the sememe depth to improve the performance of semantic similarity using different factors. Following above, the paper finishes the construction of the experimental system of sentiment analysis for short-texts.3) Implementing the system of sentiment analysis for microblog. The paper makes use the microblog messages as data resources to implement the system of sentiment analysis for microblog and gets good performance.