Genre Identification and the Compositional Effect of Genre in Literature

Proceedings of the 27th International Conference on Computational Linguistics, pages 1963–1973 Santa Fe, New Mexico, USA, August 20-26, 2018. 1963 Genre Identification and the Compositional Effect of Genre in Literature Joseph Worsham Department of Computer Science University of Colorado Colorado Springs 1420 Austin Bluffs Prkwy Colorado Springs, CO 80918, USA jworsha2@uccs.edu Jugal Kalita Department of Computer Science University of Colorado Colorado Springs 1420 Austin Bluffs Prkwy Colorado Springs, CO 80918, USA jkalita@uccs.edu Abstract Recent advances in Natural Language Processing are finding ways to place an emphasis on the hierarchical nature of text instead of representing language as a flat sequence or unordered collec- tion of words or letters. A human reader must capture multiple levels of abstraction and meaning in order to formulate an understanding of a document. In this paper, we address the problem of developing approaches which are capable of working with extremely large and complex lit- erary documents to perform Genre Identification. The task is to assign the literary classification to a full-length book belonging to a corpus of literature, where the works on average are well over 200,000 words long and genre is an abstract thematic concept. We introduce the Gutenberg Dataset for Genre Identification. Additionally, we present a study on how current deep learning models compare to traditional methods for this task. The results are presented as a baseline along with findings on how using an ensemble of chapters can significantly improve results in deep learning methods. The motivation behind the ensemble of chapters method is discussed as the compositionality of subtexts which make up a larger work and contribute to the overall genre. 1 Introduction Literary computing poses unique challenges for Natural Language Processing (NLP) and Information Retrieval (IR). Literature itself is not structured like the usual expository corpora commonly made up of Wikipedia articles or short online reviews of products and services. Rather than communicating succinct and directed information, literature is artistic and conveys complicated themes over the course of very long narratives. Given our research which shows these documents on average contain well over 200,000 words, compared to the 100-1000 words in traditional document modeling datasets, it is apparent that new techniques should be considered to effectively model documents with a much higher fidelity. This work offers a modern take on the classification problem of Genre Identification (Kessler et al., 1997). The goal of this problem is to assign the correct literary genre, such as Adventure stories, Love stories, etc., to a piece of literature using the title and body of the work. A solution to this problem would be beneficial to all organizations which supply literary services, such as libraries, online bookstores and search engines. Specifically, approaches to solve the Genre Identification problem would provide deci- sion making aids to library scientists, but could also help with recommendation systems and information retrieval. Additionally, this work builds further insight into the study of literature, themes and genre. The problem of Genre Identification also works toward a greater topic of full length literature …

Content truncated to 500 words