Effect of dependency relationships and ordered co-occurrence of words on Japanese information retrival (poster session)

ABSTRACT

We propose two Japanese information retrieval methods that enhance retrieval effectiveness using relationships between words. One is a method using dependency relationships between words in a sentence, and another is a method using the ordered co-occurrence information of words in a sentence as an approximation to the dependency relationships between them. Through retrieval experiments using the Japanese test collection for information retrieval systems NTCIR-1, we showed our two methods are superior to the TF-IDF method in retrieval effectiveness and the difference between our two methods is small. These results are independent of the document set and of the search topic set.

REFERENCES

• 1: Jose Perez-Carballo , Tomek Strzalkowski, Natural language information retrieval: progress report, Information Processing and Management: an International Journal, v.36 n.1, p.155-178, Jan. 2000 [doi>10.1016/S0306-4573(99)00049-7]
• 2: Mitra, M., Buckley, C., Singhal, A., and Cardie, C. An analysis of statistical and syntactic phrases. In Proceedings of the RIAO'97 Conference, 1997, pp. 200-214.
• 3: Matsumura, A., Takasu, A., and Adachi, J. The effect of information retrieval method using dependency relationship between words. In Proceedings of the RIAO'2000 Conference, 2000, pp. 1043-1058.

INDEX TERMS

Keywords:
co-occurence of words, dependency relationships, natural language processing, phrase