摘要
提出了一种基于隐马尔可夫模型的中文文本事件抽取方法,该方法首先通过触发词探测从文本中发现特定的候选事件语句,然后利用隐马尔可夫模型从这些语句中抽取每个候选事件的事件要素,为每一类事件要素构建一个独立的隐马尔可夫模型用于该类事件要素的抽取,构建模型的关键是模型结构的学习和参数估计。实验结果表明,该方法能较好地实现中文文本事件抽取,较其他方法有更好的抽取性能。
A method based on hidden Markov models （HMMs） is proposed for extracting the event information from Chinese texts. Firstly, the method can find a candidate sentence, which contains a description for a kind of specific event via trigger detecting. Then the method constructs a separate HMM for a kind of event argument, and makes use of these HMMs to extract event arguments from these candidate sentences. The key of constructing model is learning HMM structure and parameter estimation. Experimental results show that the method has better performance than other approaches for event extraction from Chinese texts.
出处
《微电子学与计算机》
CSCD
北大核心
2007年第10期92-94,98,共4页
Microelectronics & Computer
基金
教育部博士点基金项目(20050007023)
