Improving query translation for cross-language information retrieval using statistical models

摘要: Dictionaries have often been used for query translation in cross-language information retrieval (CLIR). However, we are faced with the problem of ambiguity, i.e. multiple translations stored a dictionary word. In addition, word-by-word is not precise enough. this paper, explore several methods to improve previous dictionary-based translation. First, as many possible, noun phrases recognized and translated whole by using statistical models phrase patterns. Second, best word selected based on cohesion words. Our experimental results TREC English-Chinese CLIR collection show that these techniques result significant improvements over simple approaches, achieve even better performance than high-quality machine system.