Title | Ensemble of RFR_SUM unigram and bigram for Chinese WSD |
Authors | Qu, Weiguang Yu, Jingsong Zhou, Junsheng Shao, Yanqiu Li, Sujian Sui, Zhifang |
Affiliation | Institute of Computational Linguistics, Peking Univ., Beijing 100871, China Department of Computer Science, Nanjing Normal Univ., Nanjing 210097, China School of Software and Microelectronics, Peking Univ., Beijing 102600, China |
Issue Date | 2007 |
Publisher | journal of computational information systems |
Citation | Journal of Computational Information Systems.2007,3,(5),1867-1874. |
Abstract | In this paper, we expand a collocation-based WSD model RFR-SUM (sum of Relative Frequency Ratio in context) from unigram (UNIRFRSUM) to bigram (BIRFRSUM) and design two algorithms for BI_RFR_SUM: Simple BI_RFR_SUM algorithm (SBI) and No Intersection BI_RFR_SUM algorithm (NI). We select 7 frequently used polysemous words as examples and the experiments show that the precision of NI algorithm can be adjusted to a very high level. We combine UNI_RFR_SUM with NI algorithm and get a precision of 96.40% with respect to that of TJNI_RFR_SUM 93-23% and SBI 93.32% in open test. This means that the ensemble learning can reduce 46.82% misclassifieation of UNIJRFRSUM model. |
URI | http://hdl.handle.net/20.500.11897/407574 |
ISSN | 15539105 |
Indexed | EI |
Appears in Collections: | 软件与微电子学院 计算语言学教育部重点实验室 |