KeisukeKamataki writeup of Poon 2007

This is a review of the paper poon_2007_joint_inference_in_information_extraction by user:KeisukeKamataki.

Summary: They applied full-joint inference approach of Markov Logic Network(MLN) for citation information identification. The method significantly outperformed previous approaches. The Joint MLN approach almost always outperformed Isolated MLN approach.

I like: This paper complements my understanding of MLN for the optional reading of this week (my main reading of this week, in a sense). This paper is informative in terms algorithm explanation and experimental methodology for using MLN.

I didn't well understand/didn't like: Although they say that the performance difference between Isolated/Joint model is statistically significant, it still seems small. And also, it would be better if they give the specific difference of computational time between Isolated/Joint model. In addition, it might be helpful if they include some examples of each joint formula.