摘要： 提出一种基于改进隐马尔可夫模型(HMM)的文本信息抽取模型。给出一个新假设，使用绝对平滑算法对模型参数进行平滑，利用Viterbi算法对观察值序列进行正序和逆序解码，基于N-Gram模型对2次解码结果进行对比消歧，得到较准确的状态序列。实验结果表 明，该信息抽取模型能提高信息抽取的准确率。

关键词: 隐马尔可夫模型, 绝对平滑, 观察值, 信息抽取, 引文信息

Abstract: This paper proposes a text information extraction model based on improved Hidden Markov Model(HMM). It gives a new assumption of observation emission. And the absolute smoothing algorithm is used to smooth the parameters of the model. The model recovers the most-likely state sequence of the observation sequence and the reverse observation sequence with the Viterbi algorithm. It compares the results with each other based on N-Gram model, and outputs a more accurate result for the state sequence. Experimental results indicate that this model has effectively improved precision.

Key words: Hidden Markov Model(HMM), absolute smoothing, observation, information extraction, citation information

中图分类号: 

• TP311