Cluster-based Unsupervised Automatic Keyphrases Extraction algorithms: experimentations on Cultural Heritage datasets

Abstract

Automatic keyword extraction is the process of identifying key terms and key phrases from documents that can appropriately represent the subject of the documents. We present here a work-in-progress, an experimentation done on unsupervised keyword extraction, with the aim of automatically associating scored keyphrases to texts, using (standard or innovative) cluster based methods, and integrating word embedding to enhance semantic relatedness of keyphrases. In the paper we present the datasets used, the state-of-theart for unsupervised automatic extraction algorithms, based on cluster methods, and we describe in details the algorithms implemented and preliminary results obtained. The results obtained are discussed, commented, and compared with those obtained, in previous experimentations, using TextRank, RAKE and Tf-idf.