With the development of the modern service industry, some problems are also created by prosperity. It is difficult to sufficiently match the supply and demand of science and technology service platforms, which leads to poor user experience. This paper puts forward a method of extracting users’ STS requirements based on improved TF-IDF and sentiment analysis. Firstly we identify user keywords by text cutting, then we calculate user requirement weight by TF-IDF. Considering semantics, we use Word2Vec to screen synonyms in to improve the accuracy of TF-IDF. Finally, the weight of user requirements is modified by sentiment analysis. The experiment demonstrates the efficiency of our method.