@techreport{oai:ipsj.ixsq.nii.ac.jp:00062363,
 author = {NGUYENPhamThanhThao and 林, 貴宏 and 尾内, 理紀夫 and 西岡, 悠平 and 竹中, 孝真 and 森, 正弥 and NGUYEN, PhamThanhThao and Takahiro, Hayashi and Rikio, Onai and Yuhei, Nishioka and Takamasa, Takenaka and Masaya, Mori},
 issue = {21},
 month = {May},
 note = {本研究は係り受け関係と相互情報量を基づき、学習不要な名詞句のカテゴリ分類手法を提案する．分類システムの入力は、各カテゴリに対する少数の種語群と分析用のコーパスのみで、学習データを必要としない．本手法は、すべての名詞句を扱うため、出現頻度が低い名詞句も分類可能である．そして、ユーザが設定した各カテゴリ（目的カテゴリと呼ぶ）以外に、ゴミカテゴリを設定する．これにより、目的カテゴリに分類すべきでない名詞句がゴミカテゴリに分類され、誤分類防止の効果があることを実験で確認した．, We propose a noun phrase categorization method without requirement for a learning phase. Our method bases on the combination of predicate-argument relations and mutual information easure8). The system input requires only a small set of seed words for each category and a text corpus for analysis, but not any learning data. We treat all noun phrases as category candidates; therefore even words with low frequency can be categorized. Also, beside the purpose categories set up by users, by setting an extra “trash category”, we could gather unexpected words into this “trash category” properly (unexpected words refer to words should not be categorized into one of the purpose categories). The experiment results showed that “trash category” is effective at reventing unexpected words from being mis - categorized into purpose categories.},
 title = {学習不要な名詞句のカテゴリ分類手法},
 year = {2009}
}