Chinese Word Segmentation based on Conditional Random Field

カテゴリ: 部門大会

論文No: GS13-3

グループ名: 【C】平成21年電気学会電子・情報・システム部門大会講演論文集

発行日: 2009/09/03

タイトル(英語): Chinese Word Segmentation based on Conditional Random Field

著者名: 于雷 (徳島大学)

著者名(英語): Yu／lei (The University of Tokushima)

キーワード: NLP|word segmentation|CRF

要約(日本語): Conditional Random Field (CRF) model is widely used in Natural Language Processing (NLP) fileld such as text classification, text summarization and Q&A system. In this paper, we use CRF model to construct Chinese word segmentation system. For CRF model in this study, several components are estimating in details, one is training template, the second is quantity of training data, the third one is training parameter. In the experiments, we use different quantity of data to train and test the models, and then the best model is confirmed to compare with the other supervised model. The results show CRF model has a better performance in all tests.

PDFファイルサイズ: 1,837 Kバイト