Research Area:  Machine Learning
Keyword or keyphrase extraction is to identify words or phrases presenting the main topics of a document. This paper proposes the Attention Rank, a hybrid attention model, to identify keyphrases from a document in an unsupervised manner. Attention Rank calculates self-attention and cross-attention using a pre-trained language model. The self-attention is designed to determine the importance of a candidate within the context of a sentence. The cross-attention is calculated to identify the semantic relevance between a candidate and sentences within a document. We evaluate the Attention Rank on three publicly available datasets against seven baselines. The results show that the Attention Rank is an effective and robust unsupervised keyphrase extraction model on both long and short documents.
Keywords:  
Keyphrase extraction
Pre-trained language model
AttentionRank
Machine Learning
Deep Learning
Author(s) Name:   Haoran Ding, Xiao Luo
Journal name:  
Conferrence name:  Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
Publisher name:  Association for Computational Linguistics
DOI:  10.18653/v1/2021.emnlp-main.146
Volume Information:  
Paper Link:   https://aclanthology.org/2021.emnlp-main.146/