- 期刊
Research on Automatic Summarization based on the Fusion of TextRank and Multi-dimensional Semantic Features
摘要
In today's Internet age, a large amount of news appears in people's daily life, and the automatic text summarization can summarize the key information and subject content of news, and thus help people reduce reading time. However, the core information of the summary generated by the traditional graph model algorithm and missing text cannot fully reflect the content of the article, so this paper proposes the automatic summary model MD-TextRank (Multi-dimensional TextRank) based on the fusion of TextRank with multi-dimensional semantic features. The method introduces word2vec to represent the news text information on the traditional TextRank algorithm, and updates the weight of sentenes from subject similarity, sentence and title similarity, keyword coverage, and feature words; and taking the military field as an example, the domain dictionary has been designed to make the generated summary better reflects the concern of the field. Experiments demonstrate that the MD-TextRank text summarization model proposed in this paper has improved 8.9,6.8, and 5.9 percentage points compared to the traditional algorithm TextRank on ROUGE-1, ROUGE-2, ROUGEL, significantly improving the quality of automatic text summarization.