Multilingual lexical resources to detect cognates in non-aligned texts

Multilingual lexical resources to detect cognates in non-aligned texts Haoxing Wang Queensland University of Technology 2 George St, Brisbane, 4000 haoxing.wang@hdr.qut.edu.au Laurianne Sitbon Queensland University of Technology 2 George St, Brisbane, 4000 l.sitbon@qut.edu.au Abstract The identification of cognates between two distinct languages has recently start- ed to attract the attention of NLP re- search, but there has been little research into using semantic evidence to detect cognates. The approach presented in this paper aims to detect English-French cog- nates within monolingual texts (texts that are not accompanied by aligned translat- ed equivalents), by integrating word shape similarity approaches with word sense disambiguation techniques in order to account for context. Our implementa- tion is based on BabelNet, a semantic network that incorporates a multilingual encyclopedic dictionary. Our approach is evaluated on two manually annotated da- tasets. The first one shows that across different types of natural text, our method can identify the cognates with an overall accuracy of 80%. The second one, con- sisting of control sentences with semi- cognates acting as either true cognates or false friends, shows that our method can identify 80% of semi-cognates acting as cognates but also identifies 75% of the semi-cognates acting as false friends. 1 Introduction Estimating the difficulty of a text for non-native speakers, or learners of a second language, is gaining interest in natural language processing, information retrieval and education communities. So far measures and models to predict readability in a cross-lingual context have used mainly text features used in monolingual readability contexts (such as word shapes, grammatical features and individual word frequency features). These fea- tures have been tested when estimating readabil- ity levels for K-12 (primary to high school) read- ers. As word frequency can be partially estimated by word length, it remains the principal feature for estimation second language learners with the assumption that less frequent words are less like- ly to have been encountered and therefore acces- sible in the memory of the learnt language by readers. However such features are not entirely adapted to existing reading abilities of learners with a different language background. In particu- lar, for a number of reasons, many languages have identical words in their vocabulary, which opens a secondary access to the meaning of such words as an alternative to memory in the learnt language. For example, English and French lan- guages belong to different branches of the Indo- European family of languages, and additionally European history of mixed cultures has led their vocabularies to share a great number of similar and identical words. These words are called cog- nates. Cognates have often slightly changed their or- thography (especially in derived forms), and quite often meaning as well in the years and cen- turies following the transfer. Because of these changes, cognates are generally of one of three different types. First of all, true cognates are English-French word pairs that are viewed as similar and are mutual translations. The spelling could be identical or not, e.g., prison and prison, ceramic and céramique. …

Content truncated to 500 words