Multi-label Text Classification Based on Improved Seq2Seq

In the era of big data, we urgently need efficient text processing technology to cope with massive amounts of text data. This paper proposes a multi label text classification method based on Sequence to Sequence (Seq2Seq), which extracts statistical features of text through Variational Auto Encoder (VAE) and integrates them into semantic features. The relationship between multiple labels is modeled as a sequence to sequence learning problem. This paper uses the Arxiv Academic Paper Dataset (AAPD) data for experiments and compares our approach with multiple benchmark models. The result indicates that, Our method outperforms traditional methods in both accuracy and average accuracy, and performs well in learning relationships between multiple labels. The results of this paper promise to provide an effective solution to the problem of multi-label text classification and lay the foundation for further optimization of Seq2Seq models in the field of natural language processing.