A review on conditional random fields as a sequential classifier in machine learning

Abstract

In this paper we present a comprehensive review of a well-known sequential classifier in machine learning Conditional Random Fields (CRFs). CRFs is proposed to cope the limitation of both generative Hidden Markov Models (HMMs) and discriminative Maximum Entropy Markov Models (MEMMs) for solving the sequential classification problems. CRFs is widely used to accomplish the sequential classification which has a temporal dimension. On its way, CRFs has been improved both on the structural learning model as well as on the area of implementation. Those areas are varying from information extraction, image understanding, computer vision, behavioral analysis, natural language processing, bioinformatics, etc. This review provides a compact and informative summary of the major research on CRFs. We present a brief description about CRFs fundamental, CRFs roadmap, and CRFs related area of implementation from several literature papers on CRFs. The contribution of this paper is to explore the roadmap of CRFs research and potential prospect in developing CRFs to solve machine learning problems, particularly problems with sequential structures.