Stylometry in a Bilingual Setup

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 977–984 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 977 Stylometry in a Bilingual Setup Silvie Cinková*, Jan Rybicki *Charles University – Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics Jagiellonian University – Institute of English Studies *Malostranské nám. 25, Praha 1, 118 00 Czech Republic 31-120 Kraków, al. Mickiewicza 9A, Poland cinkova@ufal.mff.cuni.cz, jan.rybicki@uj.edu.pl Abstract The method of stylometry by most frequent words does not allow direct comparison of original texts and their translations, i.e. across languages. For instance, in a bilingual Czech-German text collection containing parallel texts (originals and translations in both directions, along with Czech and German translations from other languages), authors would not cluster across languages, since frequency word lists for any Czech texts are obviously going to be more similar to each other than to a German text, and the other way round. We have tried to come up with an interlingua that would remove the language-specific features and possibly keep the linguistically independent features of individual author signal, if they exist. We have tagged, lemmatized, and parsed each language counterpart with the corresponding language model in UDPipe, which provides a linguistic markup that is cross-lingual to a significant extent. We stripped the output of language-dependent items, but that alone did not help much. As a next step, we transformed the lemmas of both language counterparts into shared pseudolemmas based on a very crude Czech-German glossary, with a 95.6% success. We show that, for stylometric methods based on the most frequent words, we can do without translations. Keywords: stylometry, multilinguality, Universal Dependencies; authorship attribution; translation 1. Introduction 1.1 In Search for Individual Stylistic Profiles Computational stylistics, or stylometry, is concerned with the quantitative characteristics of individual author style and the comparison of different authors, a facet of which is authorship attribution – selecting the (most likely) author of a given document from a group of authorship candidates by comparison of that document with documents by all authorship candidates. The assumption is that authors display their individual unconscious patterns of language use, and that these patterns remain constant, no matter the topic. These patterns manifest themselves in the most frequent words (mostly function words). Normally, the intra-authorial style variation should be smaller than the variation between different authors. In the stylometric comparison, the texts are compared as feature vectors of the most frequent words, and their similarity is computed as the distance of each text to each other text. The classic metric of authorial difference is Burrows’ Delta (Burrows 2002; Hoover 2004), defined as the Manhattan distance of z-scores of the frequencies of n most frequent words in the collection. Evert et al. (Evert et al., 2017) scrutinized Burrows’ Delta as well as its different modifications. They replaced the Manhattan distance in Burrows’ Delta by the cosine distance as a way of vector normalization, which substantially increased its performance. They argue that vector normalization makes the …

Content truncated to 500 words