Word Sense Disambiguation Based on Latent Maximum Entropy Model
In this paper, the latent maximum entropy model is used for word sense disambiguation of natural language processing. The major problem of current maximum entropy approach is that it can only handle with the explicit features. The latent maximum entropy model will consider the latent features such as: semantic information feature, syntactic structure feature and lexical information feature which we cant observe directly. Firstly, we show the problem formulation both for maximum entropy model and latent maximum entropy model. Then, we present how to use this new model to different kinds of features. Finally, through the improved model and the experiment for this new model, we get a conclusion. We use People Daily News datasets to test our model; the demonstration shows that our new model leads to less error and better results that the original method.
WSD latent mwimum entropy latent features exlicit features
Wenjie Su Yangsen Zhang Haiyan Kang
Institute of Intelligent Information Processing,Beijing Information Science & Technology University, Beijing Information Science & Technology University,Beijing, China
国际会议
成都
英文
122-126
2010-12-17（万方平台首次上网日期，不代表论文的发表时间）