You’ll Take the High Road and I’ll Take the Low Road: Using a Third Language to Improve Bilingual Word Alignment

You’ll Take the High Road and I’ll Take the Low Road: Using a Third Language to Improve Bilingual Word Alignment Lars Borin Department of Linguistics, Uppsala University Box 527 SE-751 20 Uppsala, Sweden Lars.Borin@ling.uu.se Abstract While language-independent sentence align- ment programs typically achieve a recall in the 90 percent range, the same cannot be said about word alignment systems, where normal recall figures tend to fall somewhere between 20 and 40 percent, in the language-independent case. As words (and phrases) for various reasons are more interesting to align than sentences, we need methods to increase word alignment recall, preferably without sacrificing precision. This paper reports on a series of experiments with pivot alignment, which is the use of one or more additional languages to improve bilingual word aligment. The conclusion is that in a multilingual parallel corpus, pivot alignment is a safe way to increase word alignment recall without lowering the precision. 1 Introduction For about a decade and a half now, researchers in Natural language processing (NLP) and general and applied linguistics have been working with parallel corpora, i.e., in the prototypical case corpora consisting of original texts in some source language (SL) together with their translations into one or more target languages (TL). In general linguistics, they are used—in the same fashion as monolingual corpora—as handy sources of authentic language. In computational linguistics and language engineering, various methods for (semi-)automatic extraction from such corpora of, among others, translation equivalents, have been explored. 2 Why is word alignment more interesting and why is it difficult? Alignment—the explicit linking of items in the SL and TL texts judged to correspond to each other—is a prerequisite for the extraction of translation equivalents from parallel corpora, and the granularity of the alignment naturally determines what kind of translation units you can get out of these resources. With sentence alignment, you get data which can be used in, e.g., translation memories. If you want to build bi- or multilingual lexica for machine translation systems (or for people), however, you want to be able to align parallel texts on the word (and phrase) level. This is because, in the last two decades, NLP grammars have become increasingly lexicalized, and grammars for machine translation—as opposed to translation memories, or example-based machine translation, neither of which uses a grammar in any interesting sense of the word—form no exception in this regard. The entries of the lexicon, which is the major repository of linguistic knowledge in a lexicalized grammar, are mainly made up of units on the linguistic levels of words and phrases. isThe problem here is that sentence alignment a fairly well-understood problem, but word alignment is much less so. This means that while language-independent sentence alignment programs typically achieve a recall in the 90 percent range, the same cannot be said about word alignment systems, where normal recall figures tend to fall somewhere between 20 and 40 percent, in the language-independent case. Thus, we need methods to increase word alignment recall, preferably without sacrificing precision.1 …

Content truncated to 500 words