[Corpora-List] Can LSI be used in Text Categorization?

[Magnus Sahlgren]

Dear Chen,

We have used Random Indexing (which is an alternative vector space
methodology that does not use SVD) to improve the performance of text
categorization. You can read about our experiments in our COLING'04 paper
that is available at:

http://www.sics.se/~mange/papers/coling2004.pdf

All the best,

/m

...............................................
         M a g n u s   S a h l g r e n
                 mange at sics.se
  SICS, Swedish Institute of Computer Science
       Box 1263, SE-164 29 Kista, Sweden
            Tel: +46 (0)8 633 16 04
	    Fax: +46 (0)8 751 72 30
        Web: http://www.sics.se/~mange/
...............................................

On Tue, 12 Oct 2004, chen wenliang wrote:

> Dear colleagues,
> 	Recently, I apply LSI to Text Categorization. I evaluate it in 20newsgroup and a Chinese Corpus. But the experimental results shows that LSI can not improve the performance.
>
>     I wonder if you could recommend me a paper which has applied LSI in Text Categorization and provides good performance. Or recomend me a good tool.
> ¡¡¡¡
>
> Best regards,
>
> Chen Wenliang chenwl at mail.neu.edu.cn
>
> Nlplab, Northeastern University, China.
>
> 									2004-10-12
>
>
>
>