Monotone statistical translation using word groups

MONOTONE STATISTICAL TRANSLATION USING WORD GROUPS Jesús Tomás1, Francisco Casacuberta2 1 Escuela Politécnica Superior de Gandia Universidad Politécnica de Valencia. 46730 Gandia, SPAIN jtomas@upv.es 2 Instituto Tecnològic d’Informàtica Universidad Politécnica de Valencia. 46071 Valencia, SPAIN fcn@iti.upv.es Abstract A new system for statistical natural language translation for languages with similar grammar is introduced. Specifically, it can be used with Romanic Languages, such as French, Spanish or Catalan. The statistical translation uses two sources of information: a language model and a translation model. The language model used is a standard trigram model. A new approach is defined in the translation model. The two main properties of the translation model are: the translation probabilities are computed between groups of words and the alignment between those groups is monotone. That is, the order between the word groups in the source sentence is conserved in the target sentence. Once, the translation model has been defined, we present an algorithm to infer its parameters from training samples. The translation process is carried out with an efficient algorithm based on stack-decoding. Finally, we present some translation results from Catalan to Spanish and compare our model with other conventional models. Keywords Statistical machine translation, words segments, search. Introduction Statistical methods have proven to be valuable in tasks such as automatic speech recognition and natural language processing (Bahl, 1983), and they present a new opportunity for automatic translation. However, current results in automatic translation are far from satisfactory (Onaizan et al., 1999; Tomás & Casacuberta 1999). We present an approach that attempts to apply Statistical Machine Translation two similar languages such as Spanish and Catalan. We think that approach is also applicable to other pairs of Romanic languages (French, Italian, Portuguese, etc.). Romanic languages allow for great flexibility in the word ordering a sentence. This feature is especially found in Spanish where almost any order of the sentence components can be correct. For example, there are dozens of ways in which the fist sentence can be expressed: Se requerirá una acción de la Comunidad para la puesta en práctica Una acción se requerirá de la Comunidad para la puesta en práctica Se requerirá de la Comunidad una acción para la puesta en práctica De la Comunidad se requerirá una acción para la puesta en práctica Para la puesta en práctica se requerirá una acción de la Comunidad ... It is true that the order in which a Spanish sentence is written can have slight differences in meaning, mainly by emphasizing certain elements in the sentence. Nevertheless, these nuances are not significant to the task we are attempting to carry out. The translation model we present stars from the following assumption “It is possible to translate a sentence written in one Romanic language to another Romanic language by translating word groups in a monotone way. To determine whether this statement was meaningful, we performed the following test: A set of 400 sentences was selected at random from different sources in three different Romanic languages- Spanish, Catalan and French. A human …

Content truncated to 500 words