@techreport{oai:ipsj.ixsq.nii.ac.jp:00049160,
 author = {朴, 哲済 and 李, 鐘赫 and 李, 根培 and Chul-Jae, Park and Jong-Hyeok, Lee and Geunbae, Lee},
 issue = {89(1995-NL-109)},
 month = {Sep},
 note = {本稿では，拡張CYK法に単語接続に関する統計モデルを利用した日本語の形態素解析手法を提案する．形態素解析の過程はまず，接続情報を検査し接続が可能な形態素解析結果をすべて得たのち，ヒューリスティックスを利用して優先順位を決める．我々は，接続情報表の値を確率として用いることにより接続の強度を表現し，その強度により形態素候補の優先順位を決めた．このとき用いる確率情報は，形態素解析の対象言語に関する確率モデルとして統計情報抽出機構から得られる．このような形態素解析手法を日本語を対象に約24万形態素のコーパスから接続情報を抽出し実験した結果95．2％の解析成功率を得た．本論文では確率接続表を用いた形態素解析能力と，接続の可不のみを表した接続表を用いた形態素解析機構を比較し評価を行う．, This paper proposes a method for Japanese morphological analysis based on the CYK algorithm using a statistical model about the conjunctive relationships between words in a sentence. The morphological analysis process is comprised of two main stages: The extraction of the several analysis results each of which are checked for morphological connectivity, and the computation of an analysis score of each morpheme chain for the results of the first stage. Through the application of statistical probabilities in connectivity information between neibouring morphemes, the most preferable analysis among the several possible ones can be selected. Probability information is extracted from a corpus of about 240,000 morphemes. Based upon these concepts, we developed a Japanese Morphological analyzer, and obtained 95.2% of accuracy for Japanese sample sentences from news. By comparing experimentally with other algorithm which represent connectivity information by bits, we demonstrate that the statistical method is more efficient.},
 title = {統計モデルによる日本語の形態素解析手法},
 year = {1995}
}