Machine translation is often times a core component of cross-lingual information retrieval systems. However, machine translation systems give poor performance for their large cost. In this paper, a machine translation system is proposedfor cross-lingual information retrieval that alleviates the large cost of standard machine translation systems while increasing performance. The system combines multiple web based translation services, which allows for a low cost solution to the translation problem. By examining the translations of multiple systems, better translation can be done. The system uses heuristics to decide if the best translation should be chosen by picking a consensus translation or using statistical analysis. Experimental results show that the proposed method is able to produce better translations than the individual systems that are used to create candidate translations.
展开▼
