Related Experiment Videos
Experiments in cross-language medical information retrieval using a mixing translation module
Tuan D Tran1, Nicolas Garcelon, Anita Burgun
1Laboratoire d'Informatique Médicale, Faculté de Médicine Université de Rennes 1, France. duc-tuan.tran@univ-rennes1.fr
Studies in Health Technology and Informatics
|September 14, 2004
Summary
This study introduces a French/English cross-language medical information retrieval (CLMIR) system. A novel mixing translation model significantly improves English document retrieval for French speakers compared to separate methods.
Area of Science:
- Medical Informatics
- Natural Language Processing
- Information Retrieval
Background:
- The vast and growing English medical literature presents challenges for non-native speakers.
- Effective cross-language information retrieval (CLIR) is crucial for accessing this information.
- Users may struggle to formulate effective English queries despite reading comprehension.
Purpose of the Study:
- To develop and evaluate a French/English CLIR system.
- To support users in retrieving English medical documents when they have difficulty with English query formulation.
- To improve the performance of cross-language medical information retrieval.
Main Methods:
- A query translation approach was employed.
- A hybrid machine translation (MT) system combining pattern-based and rule-based modules was developed.
- A multilingual UMLS Metathesaurus was used as a complementary translation resource.
Main Results:
- The proposed mixing translation module demonstrated superior performance.
- The hybrid MT approach outperformed traditional MT-only methods.
- The thesaurus-based method alone was less effective than the combined approach.
Conclusions:
- A combined, hybrid translation strategy is highly effective for French/English CLIR.
- This approach enhances the accessibility of English medical literature for French-speaking users.
- The mixing model offers a significant advancement in cross-language medical information retrieval.