Research on gender prediction based on short texts on Chinese social platforms

With the rapid development of Internet technology, various social networking platforms are becoming more and more popular. When using social platforms, users will generate a large amount of data containing rich information and value. Gender is one of the basic attributes of users. The correct gender attribute has an important influence on the research of personalized recommendation, precision marketing and user behavior. However, there are some problems in the short text of social platforms, such as scarcity of data and high noise interference, which will have a significant impact on the accuracy of prediction. This paper takes users of Sina Weibo as the research object, analyzes the emotional expressions of users of different genders, and proposes a de-noising algorithm based on emotional dictionaries to enhance the gender characteristics of users in the text and reduce the interference of noise. Experiments show that the short text can accurately predict the user's gender and improve its influence in the main evaluation indicators.