Rake-Pmi Automated Keyphrase Extraction
Somya Gupta, Namita Mittal, Alok Kumar · 2016
Keyphrase extraction is a major step which is used in various applications such as document clustering, summarization. It can be solved using supervised as well as unsupervised approach. The unsupervised approach is based on the ranking of keyphrases and extracting the top few. For ranking, researchers either use statistical measures or language modeling. We propose a system called RAKE-PMI, which uses both approaches for ranking of phrases. PointWise Mutual Information(PMI) score of adjacent words in a phrase is calculated through the language modeling. For calculation of PMI, we extract the scores from keywords extracted from the entire corpora. This helps in analyzing the term collocation of words within a phrase. We compared proposed approach with one of the popular systems called RAKE and achieved a precision of .41 as compared to the precision of .36 achieved by RAKE for the given data set.