最大熵模型在最长地点实体识别中的应用

Application of Maximum Entropy Model in the LLE Identification

实体识别是自然语言处理领域中一个十分重要的问题,是信息提取的基础,其识别程度直接影响了后续的句法分析、篇章理解等工作的精确程度。"熵"最初是热力学的一个概念,用来表示不确定度,熵越大,不确定性越大。"最大熵"模型是一种融合多种特征于一体,并综合这些特征进行建模,在满足约束的模型中选择熵最大的模型。"最大熵"模型可以综合观察到各种相关或不相关的概率知识,对许多问题的处理都可以达到较好的效果。通过实验分析了在新闻报道领域最长地点实体的特征,并应用了最大熵模型进行了识别研究。

Entity identification is an important field in Natural Language Processing （NLP）. It＇s foundation of intbrmation extracuon and its accuracy has a direct effect on many NLP tasks such as Syntactic analysis, reading comprehension and so on. ＂Entropy＂ was a concept of thermodynamics originally, used to represent uncertainty, and it decreases with increasing of uncertainty. A Maximum Entropy Model combines variety of features to model and fits the right model from all satisfied constraint model. It can observe a variety of related or not related to the probability, and can better problem solving. This paper statistically analyse characteristics of the LLE and identify using the maximum entropy model.