Abstract

Although many high-quality dictionaries contain a sufficient number of idioms for their intended users, the methods available for looking up entries in both paper and electronic dictionaries as well as in machine translation systems are not satisfactory. Providing an adequate automatic look-up function is complicated by the existence of idiom variants, which sometimes can be very creative. The problem is further complicated by the fact that the possible range of idiom variations has not been described in a computationally tractable way. Against this backdrop, we analysed the variation patterns of idioms using manually-created idiom variation data, and, on the basis of that, developed an idiom look-up system that automatically matches idiom variants in English texts with the canonical forms of idiom entries in dictionaries. The experimental results showed our system performs sufficiently well to be used in real-world settings, including as an aid for translators, which is our overall aim.