CiteULike is a free online bibliography manager. Register
and you can start organising your references online.
A Cascaded Linear Model for Joint Chinese Word Segmentation and Part-of-Speech Tagging(June 2008), pp. 897-904.
|
Reviews
[Write a review of this article]
This cascade linear model is actually a linear combination of different models, including a perceptron based on lexical-targets, a word LM, a POS LM, a co-occurrence model, and word count penalty. The training of the cascade system is in two steps, rather than training all the parameters at a time, which would be intractable. The first step is the perceptron training (of course LMs all need to be trained). And the second step is to optimizing the combination weights.
Using this method, the word segmentation and POS tagging are performed together, and the results are improved. This is sensible since joint segmentation and tagging can help each other.
- 2008-07-01 00:10:55
Find related articles from these CiteULike users Find related articles with these CiteULike tags
BibTeX record RIS record