Research on extracting online car comment tags based on sentiment analysis

In recent years, with the rapid economic growth, cars have entered thousands of households, and car review websites have emerged. After preliminary understanding, these car reviews contain the user's evaluation of various key words of the car. In this paper, a large number of car review data will be studied, including two aspects. First, it analyzes the emotional polarity of automobile comment text, so that businesses can understand users' emotional tendency to products. Secondly, on the basis of identifying the emotional polarity, the keywords concerned by users are extracted, and the satisfied and dissatisfied aspects of consumers are analyzed, which is convenient for businesses to improve in the subsequent product update and development .In this paper, crawler technology is used to collect tens of thousands of different types of car reviews on the Pacific Auto website .Research methods through natural language processing technology, text preprocessing, feature extraction and other preparatory work, for the subsequent emotional polarity analysis and keyword extraction analysis to do a good job . In the stage of emotional polarity analysis, two different models of SVM and XGBOOST are selected for comparative analysis, and a more accurate model is obtained. Then, based on the judgment of emotional polarity, lad model is selected to extract keywords.