Toward Better Loanword Identification in Uyghur Using Cross-lingual Word Embeddings

Proceedings of the 27th International Conference on Computational Linguistics, pages 3027–3037 Santa Fe, New Mexico, USA, August 20-26, 2018. 3027 Toward Better Loanword Identification in Uyghur Using Cross-lingual Word Embeddings Chenggang Mi†‡, Yating Yang†‡, Lei Wang†‡, Xi Zhou†‡, Tonghai Jiang†‡ †Xinjiang Technical Institute of Physics & Chemistry of Chinese Academy of Sciences, China ‡Key Laboratory of Speech Language Information Processing of Xinjiang, China {micg,yangyt,wanglei,zhouxi,jth}@ms.xjb.ac.cn Abstract To enrich vocabulary of low resource settings, we proposed a novel method which identify loan- words in monolingual corpora. More specifically, we first use cross-lingual word embeddings as the core feature to generate semantically related candidates based on comparable corpora and a small bilingual lexicon; then, a log-linear model which combines several shallow fea- tures such as pronunciation similarity and hybrid language model features to predict the final results. In this paper, we use Uyghur as the receipt language and try to detect loanwords in four donor languages: Arabic, Chinese, Persian and Russian. We conduct two groups of ex- periments to evaluate the effectiveness of our proposed approach: loanword identification and OOV translation in four language pairs and eight translation directions (Uyghur-Arabic, Arabic- Uyghur, Uyghur-Chinese, Chinese-Uyghur, Uyghur-Persian, Persian-Uyghur, Uyghur-Russian, and Russian-Uyghur). Experimental results on loanword identification show that our method outperforms other baseline models significantly. Neural machine translation models integrating results of loanword identification experiments achieve the best results on OOV translation(with 0.5-0.9 BLEU improvements). 1 Introduction Almost every natural language processing (NLP) task suffers from data sparseness. This situation is even worse for low resource languages in cross lingual tasks, such as neural machine transla- tion(NMT).Lexical borrowing happens in every language. It is a phenomenon of cross-linguistic in- fluence. A loanword is a word adopted from one language (the donor language) and incorporated into another language (the recipient language) without translation. One reason we choose Uyghur as the ex- ample in this study is that Uyghur has been influenced by many languages, both oriental and western languages; another reason is that Uyghur language is low-resource and lack of research in NLP field. If loanwords in a resource poor language can be identified effectively, we can use the loanword and its corresponding donor word to extend the bilingual dictionary (Tsvetkov et al., 2015). Some researchers take loanword identification as a string similarity problem. They have applied sev- eral commonly used methods to detect loanwords (Mi et al., 2014). However, these methods are usually rule-based, so these methods are difficult to deal with ambiguity. Machine learning based models have been proposed to overcome the shortcomings that exist in previous work. Lack of annotated corpora weakens the performance of these methods (Mi et al., 2016). Although some studies combine rule based and machine learning based methods to overcome data sparsity to a certain extent, challenges still exist due to semantic issues. A loanword usually shares the same/similar semantic space with its corresponding source word in donor language. Word embeddings were first proposed to learn representations where words that have the same meaning have a similar representation (Mikolov …

Content truncated to 500 words