A classification approach for detecting cross-lingual biomedical term translations

Abstract

Finding translations for technical terms is an important problem in machine translation. In particular, in highly specialized domains such as biology or medicine, it is difficult to find bilingual experts to annotate sufficient cross-lingual texts in order to train machine translation systems. Moreover, new terms are constantly being generated in the biomedical community, which makes it difficult to keep the translation dictionaries up to date for all language pairs of interest. Given a biomedical term in one language (source language), we propose a method for detecting its translations in a different language (target language). Specifically, we train a binary classifier to determine whether two biomedical terms written in two languages are translations. Training such a classifier is often complicated due to the lack of common features between the source and target languages. We propose several feature space concatenation methods to successfully overcome this problem. Moreover, we study the effectiveness of contextual and character n-gram features for detecting term translations. Experiments conducted using a standard dataset for biomedical term translation show that the proposed method outperforms several competitive baseline methods in terms of mean average precision and top-k translation accuracy.

Bilingual lexical extraction based on word alignment for improving corpus search

“…In Irvine and Callison-Burch (2016), the authors performed two experiments, the first one relying on the existence of a bilingual dictionary with no parallel texts and the second one requiring only the existence of a small amount of parallel data. Bilingual lexica were compiled for different language pairs: English/French (Bouamor et al , 2012; Hakami and Bollegala, 2017; Semmar, 2018), English/Spanish (Oliver, 2017), English/Arabic (Naguib, 2016), English/Italian and English/German (Arcan et al , 2017), English/Slovene (Vintar and Fišer, 2008), English/Croatian, Latvian, Lithuanian and Romanian (Pinnis et al , 2012), English/Chinese (Xu et al , 2015; Zhang and Wu, 2012), English/Hebrew (Tsvetkov and Wintner, 2010), English/Italian (Arcan et al , 2017), Slovak/Bulgarian (Garabík and Dimitrova, 2015), Serbian/English (Krstev et al , 2018) and so on.…”

Jointly learning word embeddings using a corpus and a knowledge base

“…For instance, a line of research has recently shown that learning accurate word or term representations is an important task in the biomedical domain. For example, [ 14 – 16 ] show that it is possible to learn cross-lingual word embeddings from UMLS [ 17 ] Metathesaurus to find translations for biomedical terms. In biomedical domain, there are large scale unstructured corpora such as Medline ( https://www.nlm.nih.gov/databases/ ) corpus, which have been extensively used for text mining tasks.…”