Chinese Journal of Psychology 2009, Vol. 51, No. 4, A

Size: px

Start display at page:

Download "Chinese Journal of Psychology 2009, Vol. 51, No. 4, A"

庚登贾
4 years ago
Views:

1 Chinese Journal of Psychology 2009, Vol. 51, No. 4, A hwawei@ncu.edu.tw Landauer 2006 (1) (2) (3) NSC I Professors Walter Kintsch and Eilleen Kintsch

2 416 Landauer Latent Semantic Analysis, LSA Landauer, 2002; Landauer, Foltz, & Laham, 1998 D e e r w e s t e r Deerwester, Dumais, Furnas, Landauer, & Harshman, 1990 Singular Value Decomposition, SVD Deerwester Landauer Foltz Laham 1998 LSA LandauerGrolier Encyclopedia 60,768 term 30,473 term-to-document cooccurrence matrix SVD Dimension Reduction 300 Landauer et al., 1998 LSA cosine value LSA Landauer et al., 1998Landauer Landauer LSA Landauer et al., 1998 Foltz, Kintsch, & Landauer, 1998LSA LSA Dennis, 2006 LSA LSA LSA LSA Quesada, 2006 LSA 2005LSA LSA

3 417 LSA TASA Touchstone Applied Science Associates, Inc Dennis, LSA (1) (2) SVD (3) Quesada, ,277 type token parsing General Text Parser GTPGiles, Wo, & Berry, 2003 SVD LSA GTP Word Boundary GTP GTP GTP SVD SVD of, an, the LSA Quesada, 2006 LSA Stop list 55,303 (1) (2) 49,021 9,277 9,277 M = 964 SD = 2335 LSA Landauer & Dumais, ,277 40,463 M = 219SD = 66 49,021 40,463 SVD GTP SVD SVD

4 418 SVD SVD TASA Tedlab edu/~dr/ SVDLIBC/ SVDlibc SVDlibc Sparse Matrix Dense Matrix SVDlibcSparse Matrix SVD 49,021 40,463PIII 2.8G 1G Ram SVD SVD 40,463 LSA Landauer Dumais Tedlab SVD LSA 1

5 LSA pairwise comparison Nearest Neighbors Pairwise Comparison 2 ASBC 1 LSA 2

6 420 term-to-term Document-to-term term-to-document document-to-document Get cosine cosine 1 1 term-to-term Document Segmentation document-to-docuement term-todocument multiple Data Get cosine Download cosine result N e a r e s t Neighbors 3 Nearest Neighbors 1 Text1 Text2 Cosine term-to-term 0.74 term-to-document 0.36 docutment-to-term 0.36 docutment-to-document 0.43 Text1 Text2 Cosine term-to-term 0.53 term-to-document 0.26 docutment-to-term 0.37 docutment-to-document 0.23

7 Term Cosine Term Cosine Threshold submit cosine10

8 422 cosine 10 4 Get CharFre Segmentation Get WordFre 198 OOV LSA cosine cosine cosine.55 cosine

9 423 3 Range Number Percentage (%) items(s) (Repeated items are not shown) ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, ,,,,,,,,,,,,,, ,, ,, , OOV ,, ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, ,,,, ,,, ,, ,, , , OOV ,,,,,,,,,,,,,,,,, relative frequency Biased Balanced word association task term

10 M SD M SD N documents term-to-documents 600 term term term-to-term 415 cosine 415 term-to-documentscosine term documents F(1, 413) = p < =.037 cosine cosinef(1, 206) = p < =.129 cosine F(1, 207) =.013 p >.05 cosine F(1, 434) = 7.95 p < =.036 cosine cosine term-to-documents term-to-term

425 6 cosine term-to-documents M SD M SD N 0.137 0.145 0.081 0.

11 425 6 cosine term-to-documents M SD M SD N cosine (1) (2) (3)

12 426 7 / F(2, 406) = p < =.731 F(1, 203) = p < =.159 cosine t =.404 p >.05 cosinecosinet =7.98; 8.51, p <.05 ASBC ASBC F F(1, 202) = all ps <.01 LSA LSA LSA Human ratinglsa cosine rating cosine Min. Max. M SD Min. Max. M SD /

13 427 LSA multiple Data /3 NSC S Deerwester, S., Dumais, S. T., Furnas, G. W., Landauer, T. K., & Harshman, R. (1990). Indexing by latent semantic analysis. Journal of the American Society for Information Science, 41, Dennis, S. (2006). How to use the LSA web site. In T. K. Landauer, D. S. McNamara, S. Dennis, & W. Kintsch (Eds.), Handbook of Latent Semantic Analysis (pp ). Mahwah, NJ: Lawrence Erlbaum Associates. Foltz, P. W., Kintsch, W., & Landauer, T. K. (1998). Analysis of text coherence using latent semantic analysis. Discourse Processes, 25, Giles, J. T.,Wo, L., & Berry, M. W. (2003). GTP (general text parser) software for text mining. In H. Bozdogan (Ed.), Statistical data mining and knowledge discovery (pp ). Boca Raton, FL: CRC press. Landauer, T. K. (2002). On the computational basis of learning and cognition: Arguments from LSA. In N. Ross (Ed.), The psychology of learning and motivation (Vol.41, pp ). New York: Academic Press. Landauer, T. K. (2006). LSA as a theory of meaning. In T. K. Landauer, D. S. McNamara, S. Dennis, & W. Kintsch (Eds.), Handbook of Latent Semantic Analysis (pp. 3-34). Mahwah, NJ: Lawrence Erlbaum Associates. Landauer, T. K., Foltz, P. W., & Laham, D. (1998). Introduction to latent semantic analysis. Discourse Processes, 25, Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato s problem: The latent semantic analysis theory of the acquisition, induction, and representation of knowledge. Psychological Review, 104, Quesada, J. (2006). Creating your own LSA spaces. In T. K. Landauer, D. S. McNamara, S. Dennis, & W. Kintsch (Eds.), Handbook of Latent Semantic Analysis (pp ). Mahwah, NJ: Lawrence Erlbaum Associates.

14 428 {P 1, P 2,.P I }{M 1, M 2,.M k }n

21 435 The Construction and Validation of Chinese Semantic Space by Using Latent Semantic Analysis Ming-Lei Chen 1, Hsueh-Cheng Wang 2, and Hwa-Wei Ko 1 1 Institute of Learning and Instruction, National Central University 2 Department of Computer Science, University of Massachusetts at Boston, MA. Recently, using corpus to create a word net is a new approach in psycholinguistic study. The present study used Latent semantic analysis (LSA) to create a Chinese semantic space. The semantic relationship between words in the Chinese semantic space was estimated by taking the dot product (cosine) between two vectors. The semantic relationship between two sentences or two documents could be estimated in the same way. The results of human data indicate that the Chinese semantic space is a valid way to represent Chinese reader s world knowledge. Keywords: latent semantic analysis, semantic relation, semantic space

第一章

第一章國立台北師範學院學報, 第十四期 ( 九十年九月 )573~614 國立台北師範學院 573 國小學生的動物生殖觀熊召弟陳業勇林益興楊婷喬摘要本研究的目的是探究國小學生對於動物生殖的想法這些概念知識主要包括五方面 :( 一 ) 生殖的