Grant: Discovery Grants Program - Individual

Cross-language information retrieval (CLIR) allows a user to formulate a query in one language and to retrieve documents in one or more languages. The current research proposal concerns a unified approach of query translation and document translation with an application on morphologically complex languages such as Arabic, Chinese, French and English. The query translation approach translates queries into document languages using available linguistic resources; while the document translation approach translates the retrieved documents into the target language. This research proposal is based on four major parts and objectives, as follows: 1. Investigate on query translation using the dictionary-based approach. An investigation on the disambiguation of polysemous lexical items using word co-occurrence graphs will be pursued. Furthermore, efficient algorithms for the transliteration of cognates and named-entities using a hybrid spelling and phonetic information will be developed. 2. Design and implement efficient methods for translating large sets of retrieved documents. Three methods based on rules, statistics and their combination, respectively, will be explored. Special data such as named entities, dates, numbers and currencies will be investigated carefully with both rule-based and statistical approaches. Furthermore, linguistics information, such as syntax and morphology are needed in the pre-processing step. Arabic and Chinese languages suffer from the problem of word segmentation. 3. Continue the research on exploiting comparable corpora for bilingual dictionary enrichment and machine translation. One of the major bottlenecks in the development of portable and robust NLP applications with broad coverage is the lack of linguistic resources. 4. Develop efficient and comparative algorithms for query expansion using an ontology-based approach and relevance feedback in both pre- and post-translation phases.