Comparison of word-based and syllable-based retrieval for Tibetan (poster session)

ABSTRACT

Tibetan retrieval based on automatically segmented words is compared with the use of overlapping syllable n-grams using a known-item retrieval evaluation. The optimal span of fixed-length n-grams is found to be 2 syllables, and indexing words is found to be as effective as indexing syllable bigrams.

REFERENCES

• 1: ACIP, Asian Classics Input Project, Release 4, 1998.
• 2: Carbonell, J., Y. Yang, R. Frederking, R.D. Brown, Y. Geng, and D. Lee, Translingual Information Retrieval: A comparative evaluation. In International Joint Conference on Artificial lntelligence, 1997.
• 3: Garofolo, J., E.M. Voorhees, V.M. Stanford, and K. Sparck-Jones, TR.EC-6 1997 Spoken Document Retrieval Track Overview and Results. In Proceedings of the Sixth Text Retrieval Conference, Gaithersburg, 1998, pp.83-91, http://trec.nist.gov
• 4: HackeR, P.G., Approaches to Tibetan Information Retrieval: Segmentation vs. n-grams. Master's Thesis. College of Library and Information Services, University of Maryland, College Park, 2000. http://www.glue.umd.edu/-oard
• 5: Miller, E., D. Shen, J. Liu, and C. Nicholas, Performance and Scalability of a Large-scale N-gram Based Information Retrieval System. Journal of Digital Information, January 2000.
• 6: Wilkenson, R., Chinese Document Retrieval at TR.EC-6. In Proceedings of the Sixth Text Retrieval Conference, Gaithersburg, 1998, pp.25-30.
• 7: Wilson, Joe, Translating Buddhism from Tibetan. Ithaca: Snow Lion Publ. 1992.

INDEX TERMS

Keywords:
Tibetan, n-gram indexing, ranked retrieval, word segmentation