Nagaoka University of Technology
   
 

--

Takeuchi, Kitajima, & Akamatsu (2000)

Takeuchi, H., Kitajima, M., & Akamatsu, M. (2000). Extracting knowledge from texts: On the effect of context length. The Fourth Asian Fuzzy Systems Symposium, 482-486.

 

Extracting Knowledge from Texts: On the Effect of Context Length

Recently, new statistical methods for extracting knowledge from huge corpus have been successfully demonstrated. Latent Semantic Analysis (LSA) is one of those corpus-based statistical methods for extracting knowledge from documents. In this paper, we examined the context length effect in LSA, and showed that five sentences separation method gave good performance. We also examined whether LSA can be appropriately applied to Japanese language as well as English, and pointed out the parameters which affect the structure of LSA result.