Philgoo Han writeup of Sha and Pereira

From Cohen Courses
Jump to navigationJump to search

This is a review of Sha_2003_shallow_parsing_with_conditional_random_fields by user:Ironfoot.

  • CRF

- This paper applies conditional random fields on shallow parsing which is NP chunking. The biggest idea of CRF was already introduced in Lafferty(2001) and this paper succeeds the idea with new approach in training. On this point I don't get the significant role of CRF in this paper becuase the actual focus seems to be on training.

  • Training

- One modified approach, two previously introduced appraoch and a baseline(general) approach is used. However, there should have been more reasoning in selecting hessian as the preconditioner in CG.

  • Result

- The accuracy of the new system using CRF is as good as previous but much better in training time.