[3Q5-GS-9] Natural language processing, information retrieval: Semantic similarity

[3Q5-GS-9-04] Consideration of a Method of Extracting Person Information and Relational Expressions from Synopses of Novels by Sequence Labeling

〇Yuji Oka¹, Kazuaki Ando¹ (1. Kagawa University)

Keywords:Novel, Person Information, Relational Expression, Sequence Labeling, Synopsis

Recently, we can read many novels easily because of the digitization of novels and the development of novel posting sites. As the number of novels increases, it becomes more difficult to find novels suitable for reader’s interest. In order to aim to construct a novel search system based on the person information, we consider a method of extracting person information and relation expressions from a synopsis of a fantasy novel by a sequence labeling model in this paper. From experimental results, we confirmed that the extraction performance of a BiLSTM-CRF model is an average of 6.5 points higher than that of the CRF model.