基于广义隐马尔可夫模型的网页信息抽取方法

Web information extraction based on a generalized hidden Markov model

针对网页所特有的基于版面结构的特点,利用基于视觉的网页分割算法VIPS对网页分块,得到一种新的状态转移序列,取代了传统的状态转移序列。通过二阶Markov链改进广义隐马尔可夫模型(GHMM)的状态转移和输出观测值假设条件,提出了二阶的广义隐马尔可夫模型。最后通过实验说明改进的GHMM对于网页信息抽取有很高的精确率。

Since web pages are based on the web-specific layout structure feature, instead of using the transitional sequential state transition order, a new state transition order was proposed by using a vision based page segmentation algorithm （VIPS）. In addit- ion, the supposed state transition and the emission symbol conditions were improved by using the second-order Markov chain, and then a novel generalized hidden Markov model （GHMM） was proposed based on the improvement. Finally,through an example, it shows that the modified GHMM has a very high precision for web information extraction.