Research on employment sentiment analysis of college graduates based on global and local key fusion features

The increasingly fierce competition in the job market has brought unprecedented pressure and challenges to college students who are struggling to find a job. In order to address the manual analysis insufficiency of the mental state of college graduates who are struggling to find a job, this paper proposes a method based on global and local key fusion features, in order to provide necessary manual intervention. Based on the rough set theory, the importance of employment sentiment-related words for college graduates is defined, and a local key information module is constructed to extract important words from interview data. Secondly, convolutional neural networks are used to mine the potential information regarding sentiment features of college graduates at both global and local scales. A channel attention module is introduced to enhance the model's ability to focus on sentiment features of job hunters without increasing the model size. At the same time, the pooling layer is adjusted to an overlapping pooling layer, and the network feature extraction performance is improved by combining the pooling strategy of multi-level pyramids. Finally, in order to effectively utilize the global information of the text and the local information of words, the sentiment information of the global and local regions is integrated to achieve multi-dimensional analysis of sentiment data collected by interviewing college graduates. The experimental results show that the algorithm can effectively obtain semantic information about the mood of college students counseled and accurately warn of the data reflecting negative emotion.