摘要
A kind of high-powered English BaseNP (base noun phrase) identifier which combines the strengths of statistics-based and rule-based learning is designed and implemented. This identifier views TBL(Transformation-based learning) algorithm as its whole algorithm, applying maximum entropy algorithm to its initialization process during its two phases--rule learning and final tagging. The contrastive experimental results show that this combinative method is superior to both algorithms if applied separately and achieves high precision and recall of over 94%.
源语言 | 英语 |
---|---|
页(从-至) | 647-652 |
页数 | 6 |
期刊 | Journal of Computational Information Systems |
卷 | 4 |
期 | 2 |
出版状态 | 已出版 - 4月 2008 |