Improving Translation Lexicon Induction from Monolingual Corpora via Dependency Contexts and Part-of-Speech Equivalences

Proceedings of the Thirteenth Conference on Computational Natural Language Learning (CoNLL), pages 129–137, Boulder, Colorado, June 2009. c⃝2009 Association for Computational Linguistics Improving Translation Lexicon Induction from Monolingual Corpora via Dependency Contexts and Part-of-Speech Equivalences Nikesh Garera, Chris Callison-Burch, David Yarowsky Department of Computer Science, Johns Hopkins University Baltimore MD, USA {ngarera,ccb,yarowsky}@cs.jhu.edu Abstract This paper presents novel improvements to the induction of translation lexicons from monolingual corpora using multilin- gual dependency parses. We introduce a dependency-based context model that in- corporates long-range dependencies, vari- able context sizes, and reordering. It pro- vides a 16% relative improvement over the baseline approach that uses a fixed context window of adjacent words. Its Top 10 accuracy for noun translation is higher than that of a statistical translation model trained on a Spanish-English par- allel corpus containing 100,000 sentence pairs. We generalize the evaluation to other word-types, and show that the per- formance can be increased to 18% rela- tive by preserving part-of-speech equiva- lencies during translation. 1 Introduction Recent trends in machine translation illustrate that highly accurate word and phrase translations can be learned automatically given enough parallel training data (Koehn et al., 2003; Chiang, 2007). However, large parallel corpora exist for only a small frac- tion of the world’s languages, leading to a bottleneck for building translation systems in low-density lan- guages such as Swahili, Uzbek or Punjabi. While parallel training data is uncommon for such lan- guages, more readily available resources include small translation dictionaries, comparable corpora, and large amounts of monolingual data. The marked difference in the availability of monolingual vs parallel corpora has led several researchers to develop methods for automatically learning bilingual lexicons, either by using mono- lingual corpora (Rapp, 1999; Koehn and Knight, 2002; Schafer and Yarowsky, 2002; Haghighi et al., 2008) or by exploiting the cross-language evidence of closely related “bridge” languages that have more resources (Mann and Yarowsky, 2001). This paper investigates new ways of learning translations from monolingual corpora. We extend the Rapp (1999) model of context vector projection using a seed lexicon. It is based on the intuition that translations will have similar lexical context, even in unrelated corpora. For example, in order to translate the word “airplane”, the algorithm builds a context vector which might contain terms such as “passen- gers”, “runway”, “airport”, etc. and words in tar- get language that have their translations (obtained via seed lexicon) in surrounding context can be con- sidered as likely translations. We extend the basic approach by formulating a context model that uses dependency trees. The use of dependencies has the following advantages: • Long distance dependencies allow associated words to be included in the context vector even if they fall outside of the fixed-window used in the baseline model. • Using relationships like parent and child in- stead of absolute positions alleviates problems when projecting vectors between languages with different word orders. • It achieves better performance than baseline context models across the board, and better performance than statistical translation models on Top-10 accuracy …

Content truncated to 500 words