Implementation of writing support system for translating collocations based on the word co-occurrence on the web documents

This paper explains the implementation of the system for retrieving candidates for the translation of collocations (combinations of words) from the web. For non-native speakers like Japanese, writing in English is a hard work in order not to use unnatural collocations. Sometimes they have to consult dictionaries or real documents to judge naturalness of English collocations. Besides, it is also difficult to seek out the proper candidates even using dictionaries or real documents. The proposed system is effective to seek out the proper translation of such a collocation. When a user inputs a Japanese collocation "n^J WO v^J" and gathers some nouns each of which co-occurs strongly with v^J as the same meaning as v^J in "n^J WO v^J", the system retrieves candidates for the translation of v^J from the web with their priorities. After that, the user checks validity of each candidate using the real document which the system provides and includes the English phrase consisting of a predicate verb V^E and its object n^E. Here, n^E is the translation of n^J and V^E is the candidates for the translation of v^J.