JAIST Repository >
科学技術開発戦略センター 2003~2008 >
z2-70. JAIST PRESS 発行誌等 >
IFSR 2005 >

このアイテムの引用には次の識別子を使用してください: http://hdl.handle.net/10119/3906

タイトル: Co-Training of Conditional Random Fields for Segmenting Sequence Data
著者: Xuan-Hieu, Phan
Le-Minh, Nguyen
Inoguchi, Yasushi
キーワード: semi-supervised learning
conditional random fields
text labeling and segmentation
発行日: Nov-2005
出版者: JAIST Press
抄録: This paper presents a semi-supervised co-training approach for discriminative sequential learning models, such as conditional random fields (CRFs). In this framework, different CRF models are trained on an initial set of sequence data according different views. The bootstrapping process is performed by iteratively adding new reliably inferred data sequences to the training data sets of CRF models retraining them. Reliable data sequences are inferred from a huge set of unlabeled data by estimating entropy values of predicted labels at time positions in data sequences. The inference and re-train operations are repeated a number of times in order that each CRF model should gain as much useful evidence from unlabeled data and the other CRF models as possible. The proposed method was tested on noun phrase chunking and achieved significant results.
記述: The original publication is available at JAIST Press http://www.jaist.ac.jp/library/jaist-press/index.html
IFSR 2005 : Proceedings of the First World Congress of the International Federation for Systems Research : The New Roles of Systems Sciences For a Knowledge-based Society : Nov. 14-17, 2116, Kobe, Japan
Symposium 5, Session 2 : Data/Text Mining from Large Databases Text Mining
言語: ENG
URI: http://hdl.handle.net/10119/3906
ISBN: 4-903092-02-X
出現コレクション:IFSR 2005


ファイル 記述 サイズ形式
20134.pdf147KbAdobe PDF見る/開く



お問い合わせ先 : 北陸先端科学技術大学院大学 研究推進課図書館情報係