马尔可夫逻辑网在信息抽取中的应用

Application of Markov Logical Networks in Information Extraction

提出了一个基于马尔可夫逻辑网的信息抽取方法,将所有记录的分割和记录去重在一个单独的整合推理过程中进行。由于采用马尔可夫逻辑和现有的推理算法,其主要工作是编写合适的逻辑公式,工程量比其他传统方法少得多。实验基于CiteSeer和Cora这两个引文匹配数据集,其结果要明显优于之前的其他方法,同时也证明了马尔可夫逻辑网模型的精确性。

A method based on Markov logic network is proposed for information extraction. The tasks of segmentation and de - duplication of all records are performed together in a single integrated inference process. Taking the advantage bf the Markov logic and existing reasoning al- gorithm, the major job is to carry out appropriate logical formulas, of which engineering is much less than other traditional methods. The exper- iment is performed based on the two CiteSeer and Cora datasets for citation matching. Experimental results show that the method is superior to prevenient ones,with the accuracy of MLN model proved.