| 作成者 |
|
|
|
|
|
| 本文言語 |
|
| 出版者 |
|
|
|
| 発行日 |
|
| 収録物名 |
|
| 巻 |
|
| 号 |
|
| 開始ページ |
|
| 終了ページ |
|
| 出版タイプ |
|
| アクセス権 |
|
| JaLC DOI |
|
| 関連DOI |
|
| 関連URI |
|
| 関連情報 |
|
| 概要 |
In Japanese language, derivative word consists of a noun and some following suffixes, and those components are concatenated into a sequence without separators. This construction often cause ambiguity ...in parsing or Kana-Kanji conversion. Some methods to treat derivatives have been developed; 1) recognizing arbitrary combination of any noun and any suffix, 2) registering collected derivative words directly into the word dictionary, and 3) using semantic category to enable selectional restriction. However, these methods have too simple mechanism to derive correct analysis. In our previous paper, we proposed 4) an example-based method in which collected sample words are generalized for wider coverage of derivative words. In this paper, we compare these methods through experiments. To realize fair comparison, all methods were represented in Probabilistic Context Free Grammar (PCFG) and equally tuned with the same training method Maximum Likelihood Estimate. The results show that our method is superior to the methods 1) and 3), under the condition that the grammar learned more than 80,000 generalized examples.続きを見る
|
Mendeley出力
