| Summary: | 碩士 === 國立清華大學 === 資訊工程學系 === 97 === In this paper, we introduce a new method for learning to find translation equivalents of a given collocation on the Web based on the query expansion strategy. Our approach involves finding translations in a parallel corpus and learning query expansion terms for the given collocation in order to bias search engines towards returning the top-ranked snippets containing sought-after translations. We utilized the corpus translations from parallel corpus and attempt to learn additional QE terms for retrieving more translations on the Web. The query expansion method is trained on a parallel corpus and validated on the Web. At run time, a given collocation is automatically transformed into a set of queries and sent to a search engine. Then candidate translations are retrieved from the returned snippets and ranked according to their similarity with respect to the corpus translations. Our method provides significantly more translation equivalents from the Web in addition to translations found in parallel corpus, which could be used to assist language learners, translator, and the development of machine translation systems.
|
|---|
