Cognate Production using Character-based Machine Translation

International Joint Conference on Natural Language Processing, pages 883–891, Nagoya, Japan, 14-18 October 2013. Cognate Production using Character-based Machine Translation Lisa Beinborn, Torsten Zesch and Iryna Gurevych Ubiquitous Knowledge Processing Lab (UKP-TUDA) Department of Computer Science, Technische Universit¨at Darmstadt Ubiquitous Knowledge Processing Lab (UKP-DIPF) German Institute for International Educational Research www.ukp.tu-darmstadt.de Abstract Cognates are words in different languages that are associated with each other by lan- guage learners. Thus, cognates are im- portant indicators for the prediction of the perceived difficulty of a text. We in- troduce a method for automatic cognate production using character-based machine translation. We show that our approach is able to learn production patterns from noisy training data and that it works for a wide range of language pairs. It even works across different alphabets, e.g. we obtain good results on the tested language pairs English-Russian, English-Greek, and English-Farsi. Our method performs sig- nificantly better than similarity measures used in previous work on cognates. 1 Introduction In order to improve comprehension of a text in a foreign language, learners use all possible infor- mation to make sense of an unknown word. This includes context and domain knowledge, but also knowledge from the mother tongue or any other previously acquired language. Thus, a student is more likely to understand a word if there is a sim- ilar word in a language she already knows (Ring- bom, 1992). For example, consider the following German sentence: Die internationale Konferenz zu kritischen Infrastrukturen im Februar ist eine Top- Adresse f¨ur Journalisten. Everybody who knows English might grasp the gist of the sentence with the help of associ- ated words like Konferenz-conference or Februar- February. Such pairs of associated words are called cognates. A strict definition only considers two words as cognates, if they have the same etymological ori- gin, i.e. they are genetic cognates (Crystal, 2011). Language learners usually lack the linguistic back- ground to make this distinction and will use all similar words to facilitate comprehension regard- less of the linguistic derivation. For example, the English word strange has the Italian correspondent strano. The two words have different roots and are therefore genetically unrelated. However, for lan- guage learners the similarity is more evident than for example the English-Italian genetic cognate father-padre. Therefore, we aim at identifying all words that are sufficiently similar to be associated by a language learner no matter whether they are genetic cognates. As words which are borrowed from another language without any modification (such as cappuccino) can be easily identified by direct string comparison, we focus on word pairs that do not have identical spelling. If the two associated words have the same or a closely related meaning, they are true cognates, while they are called false cognates or false friends in case they have a different meaning. On the one hand, true cognates are instrumental in construct- ing easily understandable foreign language exam- ples, especially in early stages of language learn- ing. On the other hand, false friends are known to be a …

Content truncated to 500 words