Efficient Automatic Correction of Misspelled Arabic Words Based on Contextual Information
暂无分享,去创建一个
We address in this paper a new method aiming to reduce the number of proposals given by automatic Arabic spelling correction tools. We suggest the use of error’s context in order to eliminate some correction candidates. Context will be nearby words and can be extended to all words in the text. We present here experimentations we performed to validate some hypotheses we made. Then we detail the method itself. Finally we describe the experimental protocol we used to evaluate the method’s efficiency. Our method was tested on a corpus containing genuine errors and has yielded good results. The average number of proposals has been reduced by about 75% (from 16.8 to 3.98 proposals on average).
[1] Kemal Oflazer,et al. Spelling Correction in Agglutinative Languages , 1994, ANLP.
[2] David Yarowsky,et al. Discrimination Decisions for 100,000-Dimensional Spaces , 1995 .
[3] Karen Kukich,et al. Techniques for automatically correcting words in text , 1992, CSUR.
[4] Eneko Agirre,et al. Towards a Single Proposal in Spelling Correction , 1998, COLING-ACL.