Dictionary-Based Thai CLIR: An Experimental Survey of Thai CLIR

摘要

This paper describes our participation in the Cross-Language Evaluation Forum. Our objectives for this experiment were three-fold. Firstly, the coverage of the Thai bilingual dictionary was evaluated when translating queries. Secondly, we investigated whether the segmentation process affected the CLIR. Lastly, this research examines query formation techniques. Since this is our first international experimental in CLIR, our approach used dictionary-based techniques to translate Thai queries into English queries. Four runs were submitted to CLEF: (a) single mapping translation with manual segmentation, (b) multiple mapping translation with manual segmentation, (c) single mapping translation with automatic segmentation and (d) single mapping with query enhancement using words from our Thai thesaurus. The retrieval effectiveness was worse than we expected. The simple dictionary mapping technique is unable to achieve retrieval effectiveness, although the dictionary lookup gave a very high percentage of mapping words. The words from the dictionary lookup are not specific terms but each is mapped to a definition or meaning of that term. Furthermore, Thai stopwords, stemmed words and word separation have reduced the effectiveness of Thai CLIR.