Document classification based on c

Summary form only given. This paper proposes a document classification approach based on word vectors. By learning context relationships, word vectors may represent fine-grained semantic elements. We assume that these low-level semantic implementation of words can be composed to represent high-level semantic concepts, and thus the semantic content of a document can be derived from those of the words it involves. Our experiments confirm that, even with the simplest pooling method, the document representation based on word vectors can deliver good performance on text classification tasks. When compared to the conventional LDA-based approach, the word vector approach is more stable, efficient and generalizable.