Multi-Label Text Classification Based on Multidimensional Information Extraction

At present, there is a common problem of project labels missing in open source software (OSS) communities. However, proper project labels help OSS users find the expected OSS projects more efficiently and improve the quality of project recommendations. In response to the above challenges, this paper proposes a multi-label text classification based on syntactic and semantic (MLSS), which combines semantic, syntactic, implicit topic, and text distribution, can effectively solve the problems of sparse text features and poor topic focus of open source projects, and improve the classification effect of OSS project text labels. Compared with the classical ML-kNN and BERT model, the experiment results show that the proposed MLSS algorithm has advantages in nearly all five evaluation criteria, especially the average precision is increased 16.46% higher than that of the BERT model and 103.80% higher than the ML-kNN model.