Filtering Wiktionary Triangles by Linear Mbetween Distributed Word Models

Filtering Wiktionary triangles by linear mapping between distributed word models Márton Makrai Research Institute for Linguistics Hungarian Academy of Sciences Benczúr u. 33, H-1068 Budapest makrai.marton@nytud.mta.hu Abstract Triangulation infers word translations in a pair of languages based on translations to other, typically better resourced ones called pivots. This method may introduce noise if words in the pivot are polysemous. The reliability of each triangulated translation is basically estimated by the number of pivot languages (Tanaka and Umemura, 1994). Mikolov et al. (2013b) introduce a method for scoring word translations. Translation is formalized as a linear mapping between distributed vector space models (VSM) of the two languages. VSMs are trained on monolingual data, while the mapping is learned in supervised fashion, using a seed dictionary of some thousand word pairs. We apply linear mapping to filter triangulated translations, and show that scores by the mapping are smoother measure of merit than the number of pivots. The methods we use are language-independent, and the training data is easy to obtain for many languages. We chose the German-Hungarian pair for evaluation, in which the filtered triangles resulting from our experiments are the greatest freely available list of word translations we are aware of. Keywords: word triangulation, word embedding, Wiktionary Word translations arise in dictionary-like organization as well as via machine learning from corpora. The former is exemplified by Wiktionary, a crowd-sourced dictionary with editions in many languages. Ács et al. (2013) ob- tain word translations from Wiktionary with the pivot-based method, also called triangulation, that infers word transla- tions in a pair of languages based on translations to other, typically better resourced ones called pivots. Triangulation may introduce noise if words in the pivot are polysemous. The reliability of each triangulated translation is basically estimated by the number of pivot languages (Tanaka and Umemura, 1994). Mikolov et al. (2013b) introduce a method for generating or scoring word translations. Translation is formalized as a linear mapping between distributed vector space models (VSM) of the two languages. VSMs are trained on mono- lingual data, while the mapping is learned in a supervised fashion, using a seed dictionary of some thousand word pairs. The mapping can be used to associate existing trans- lations with a real-valued similarity score. This paper exploits human labor in Wiktionary combined with distributional information in VSMs. We train VSMs on gigaword corpora, and the linear translation mapping on direct (non-triangulated) Wiktionary pairs. This mapping is used to filter triangulated translations based on scores. The motivation is that scores by the mapping may be a smoother measure of merit than considering only the number of pivot for the triangle. We evaluate the scores against dictionar- ies extracted from parallel corpora (Tiedemann, 2012). We show that linear translation really provides a more reliable method for triangle scoring than pivot count. The methods we use are language-independent, and the training data is easy to obtain for many languages. We chose the German-Hungarian pair for evaluation, in which the filtered triangles resulting from our experiments …

Content truncated to 500 words