论文信息 - Information Space Gets Normal

Information Space Gets Normal

Experiments are presented based on unofficial results for TREC-7. Eigensystems analysis of a term cooccurrence matrix is compared to eigensystems analysis of a term correlation matrix. For each matrix type, the effect of term weighting and document length normalization is assessed. Recall-precision curves and other TREC statistics indicate that the use of the correlation matrix improves performance regardless of what term weighting or document length normalization is used.

Gregory B. Newby | G. Newby

[1] Richard A. Harshman,et al. Indexing by Latent Semantic Analysis , 1990, J. Am. Soc. Inf. Sci..

[2] Venkata Subramaniam,et al. Information Retrieval: Data Structures & Algorithms , 1992 .

[3] Gregory B. Newby. Metric Multidimensional Information Space , 1996, TREC.

[4] T. Landauer,et al. Indexing by Latent Semantic Analysis , 1990 .

[5] Ricardo Baeza-Yates,et al. Information Retrieval: Data Structures and Algorithms , 1992 .

[6] Gregory B. Newby. Context-Based Statistical Sub-Spaces , 1997, TREC.

[7] Myron Wish,et al. Basic Concepts of Multidimensional Scaling , 1978 .

[8] Peter Ingwersen,et al. Cognitive Perspectives of Information Retrieval Interaction: Elements of a Cognitive IR Theory , 1996, J. Documentation.