Odi et Amo. Creating, Evaluating and Extending Sentiment Lexicons for Latin.

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3078–3086 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3078 Odi et Amo Creating, Evaluating and Extending Sentiment Lexicons for Latin Rachele Sprugnoli, Marco Passarotti, Daniela Corbetta, Andrea Peverelli CIRCSE Research Centre, Universit`a Cattolica del Sacro Cuore Largo Agostino Gemelli 1, 20123 Milano {rachele.sprugnoli,marco.passarotti}@unicatt.it Abstract Sentiment lexicons are essential for developing automatic sentiment analysis systems, but the resources currently available mostly cover modern languages. Lexicons for ancient languages are few and not evaluated with high-quality gold standards. However, the study of attitudes and emotions in ancient texts is a growing field of research which poses specific issues (e.g., lack of native speakers, limited amount of data, unusual textual genres for the sentiment analysis task, such as philosophical or documentary texts) and can have an impact on the work of scholars coming from several disciplines besides computational linguistics, e.g. historians and philologists. The work presented in this paper aims at providing the research community with a set of sentiment lexicons built by taking advantage of manually-curated resources belonging to the long tradition of Latin corpora and lexicons creation. Our interdisciplinary approach led us to release: i) two automatically generated sentiment lexicons; ii) a Gold Standard developed by two Latin language and culture experts; iii) a Silver Standard in which semantic and derivational relations are exploited so to extend the list of lexical items of the Gold Standard. In addition, the evaluation procedure is described together with a first application of the lexicons to a Latin tragedy. Keywords: sentiment analysis, Latin, evaluation. 1. Introduction The most common applications of resources and tools for sentiment analysis, i.e. the task of automatically classifying a piece of text according to the sentiment conveyed by it, fall into categories like social media and customer experi- ence monitoring, or people’s opinion mining. Accordingly, the texts which are most frequently analysed with senti- ment analysis tools, and included in linguistic resources for sentiment analysis, are the kind of tweets, comments, feed-backs and social media chats. These are also the most widespread textual typologies used for building such tools in data-driven fashion, by exploiting a wide set of empirical data for training. Said characteristics of sentiment analysis tools and re- sources (i.e. applications, textual typologies and size of data) do not take place when historical and ancient lan- guages are concerned. If available, the most frequent ap- plication of sentiment analysis tools and resources in such area would deal with evaluating the lexical properties of lit- erary, philosophical, or documentary texts, to assess their degree of positive, neutral, or negative sentiment. Unlike social media texts, which grow hourly in size, texts written in historical and ancient languages are provided by closed corpora, which very rarely can be extended thanks to newly found texts. The Open Greek and Latin project1, whose ul- timate goal is to represent every source text produced in Classical Greek or Latin during antiquity (up to c. 600 …

Content truncated to 500 words