Automatic Keyword Extraction Using TextRank

Summarizing and extracting keywords from textual documents is a fundamental task involving in many applications in natural language processing and related fields. This work presents an automatic keyword extraction algorithm based primarily on a weighted TextRank model. In this model, word embedding vectors are used to compute a similarity measure as an edge weight. Incorporating sentence importance scores derived from the TextRank model at a sentence level enhances an overall performance. The proposed algorithm is experimented and compared with the traditional TextRank algorithm as well as the weighted TextRank algorithm with word embedding-based weights.