Identifying Cognates in English-Dutch and French-Dutch by means of Orthographic Information and Cross-lingual Word Embeddings

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4096–4101 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4096 Identifying Cognates in English-Dutch and French-Dutch by means of Orthographic Information and Cross-lingual Word Embeddings Els Lefever, Sofie Labat and Pranaydeep Singh LT3, Language and Translation Technology Team, Ghent University Groot-Brittanni¨elaan 45, 9000 Ghent, Belgium {firstname.lastname}@ugent.be Abstract This paper investigates the validity of combining more traditional orthographic information with cross-lingual word embeddings to identify cognate pairs in English-Dutch and French-Dutch. In a first step, lists of potential cognate pairs in English-Dutch and French-Dutch are manually labelled. The resulting gold standard is used to train and evaluate a multi-layer perceptron that can distinguish cognates from non-cognates. Fifteen orthographic features capture string similarities between source and target words, while the cosine similarity between their word embeddings represents the semantic relation between these words. By adding domain-specific information to pretrained fastText embeddings, we are able to obtain good embeddings for words that did not yet have a pretrained embedding (e.g. Dutch compound nouns). These embeddings are then aligned in a cross-lingual vector space by exploiting their structural similarity (cf. adversarial learning). Our results indicate that although the classifier already achieves good results on the basis of orthographic information, the performance further improves by including semantic information in the form of cross-lingual word embeddings. Keywords: cognate detection, multi-layer perceptron, orthographic similarity, cross-lingual word embeddings 1. Introduction Cross-lingual similarity of word pairs from different languages concerns both formal and semantic overlap. Whereas the former refers to orthographic and/or phonetic similarity, the latter refers to translation equivalents in dif- ferent languages (Schepens et al., 2013). Cognates are then defined as word pairs in different languages that have a sim- ilar form and meaning, which is often the result of a shared linguistic origin in some ancestor language (Frunza and Inkpen, 2007). Furthermore, true cognates can be distin- guished from false friends, which have a similar form but different meaning, and from partial cognates, which share the same meaning for some, but not all contexts. For exam- ple, the English-Dutch word pair father – vader is a cog- nate pair, as both words in the pair have a similar form and meaning. In contrast, the French-Dutch gras – gras and the English-Dutch driving – drijvende are, respectively, in- stances of false friends and partial cognates. In the first word pair, the French gras means “fat, greasy”, while the Dutch gras stands for “grass”. In the second word pair, the English driving is defined as “communicating/ having great force” or “exerting pressure” by Merriam Webster1, but it is also often used in contexts such as “driving a vehicule”, while for Dutch, “drijvende” is not used in this latter con- text. Although cognates are often historically related, we chose to not incorporate this etymological criterion into the present study, as most previous research on computational cognate detection (see e.g. Schepens et al. (2013)) also dis- regards this criterion. Cognate pair lists …

Content truncated to 500 words