Bilingual lexicon induction through a pivot language

Abstract

This paper presents a new method for constructing bilingual lexicons through a pivot language. The proposed method is adapted from the context-based approach, called the standard approach, which is well-known for building bilingual lexicons using comparable corpora. The main difference between the standard approach and the proposed method is how to represent context vectors. The former is to represent context vectors in a target language, while the latter in a pivot language. The proposed method is very simplified from the standard approach thereby. Furthermore, the proposed method is more accurate than the standard approach because it uses parallel corpora instead of comparable corpora. The experiments are conducted on a language pair, Korean and Spanish. Our experimental results have shown that the proposed method is quite attractive where a parallel corpus directly between source and target languages are unavailable, but both source-pivot and pivot-target parallel corpora are available.

Evaluating a Pivot-Based Approach for Bilingual Lexicon Extraction

“…Most of them are based on the context-based approach, so called the standard approach, proposed by Rapp [ 10 ]. In this section, first of all, we describe the standard approach (SA) in short and then the pivot-based standard approach (PBSA) proposed by Kim et al [ 13 , 14 ].…”

“…As mentioned in the previous section, there are many advantages in the standard approach, but the standard approach still has some problems on resource-poor language pairs like Korean and Bengali. To overcome these problems, the PBSA was proposed by Kim et al [ 13 , 14 ], with adapting from the standard approach. Figure 1 is the overall structure of the PBSA, where s i , t i , and e i are the i th word in a source language, a target language, and a pivot language, respectively, and n , m , and l are also the number of words in a source language, a target language, and a pivot language, respectively.…”

“…The two phrase tables are extracted from two parallel corpora on the basis of phrase based statistical machine translation (SMT). The third is to build bilingual lexica based on the similarity of context vectors in a pivot language [ 13 , 14 ]. The context vectors (called source and target context vectors) in this approach are represented by words in a pivot language and are produced using two parallel corpora sharing the pivot language, that is, source-pivot and pivot-target parallel corpora.…”