| Item type |
FIT(1) |
| 公開日 |
2007-08-22 |
| タイトル |
|
|
言語 |
en |
|
タイトル |
E-063 Context Dependent Class Language Model Based on LSA |
| 言語 |
|
|
言語 |
eng |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_5794 |
|
資源タイプ |
conference paper |
| 著者所属 |
|
|
|
豊橋科技大 |
| 著者所属 |
|
|
|
豊橋科技大 |
| 著者所属 |
|
|
|
豊橋科技大 |
| 著者所属(英) |
|
|
|
en |
|
|
Toyohashi University of Technology, Department of Information and Computer Sciences |
| 著者所属(英) |
|
|
|
en |
|
|
Toyohashi University of Technology, Information and Media Center |
| 著者所属(英) |
|
|
|
en |
|
|
Toyohashi University of Technology, Department of Information and Computer Sciences |
| 著者名 |
/ 土屋, 雅稔
中川, 聖一
|
| 著者名(英) |
Naptali, Welly
Tsuchiya, Masatoshi
Nakagawa, Seiichi
|
| 論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
We propose an alternative method in language model, called context dependent class language model (CDC), to solve data sparseness problem which is suffered by n-gram language model. The proposing method makes usage of the successful ideas of latent semantic analysis (LSA) in projecting discrete words into continuous vector space. We perform classification on the resulting space and then for- mulate the CDC. Experimental results on the Wall Street Journal (WSJ) corpus show that the interpolation of the proposed method and a backoff trigram model, achieves better performance than state-of-the-art trigram language model as a baseline. |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AA11740605 |
| 書誌情報 |
情報科学技術フォーラム一般講演論文集
巻 6,
号 2,
p. 291-294,
発行日 2007-08-22
|
| 出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |
KJ00006548720.pdf (436.2 kB)
