UTACLIR @ CLEF 2002: Towards a Unified Translation Process Model

The UTACLIR query translation system was originally designed for the CLEF 2000 and 2001 campaigns. In the two first years the query translation application consisted of separate programs based on common translation principles for the language pairs Finnish English, German English and Swedish English. The idea of UTACLIR is based on recognizing distinct source key types and processing them accordingly. The linguistic resources utilized by the framework include morphological analysis or stemming in indexing, stop word removal, normalization of topic words, splitting of compounds written together, handling of non-translated words, phrase composition of compounds in the target language, bilingual dictionaries and structured queries. This year we participated in CLEF with the new UTACLIR system, which is a single program unified for all the languages. The user gives the system the codes of source and target language as well as the query to be translated. The UTACLIR system chooses the language resources upon the codes the user has given. A morphological analyser is used to process the source language words in order to match the words in the translation dictionary. We have utilized an 18-language dictionary (with all possible language pairs) as the main translation resource of the UTACLIR. It is also possible to implement other parallel dictionaries. For the target language it is possible to use either a morphological analyser or a stemmer.