Review of methodology about knowledge extraction

This paper presents a review of the latest research state of knowledge extraction. An informetric approach is adopted in which literature annual growth, prolific writers, core journals and high-frequency keywords are used as indicators. First, high-frequency keywords are classified according to their relations with concepts in knowledge extraction. Then, methods of knowledge extraction are inducted and generalized. There are two types of methods, i.e., rule-based methods and statistical methods. The former consists of model rules, syntactic rules, weight word rules, etc. while the latter includes weak supervised learning algorithm and strong supervised learning algorithm. Based on quantitative analysis of papers and qualitative analysis of methods, we are able to point out the obstacles and trends in knowledge extraction. ? 2015, ICIC International.EI82107-2115