Nlao writeup of Borkar 2001

This is a review of Borkar_2001_Automatic_Segmentation_of_Text_Into_Structured_Records by user:Nlao.

This is a very solid work with:

- interesting model: nested HMM with parallel path merging

- realistic techniques: absolute disconting, hierarchical feature selection, integrate partial database

- comprehensive evaluation: 2 tasks, 4 datasets, 3 baselines

[minor points]

- failed to evaluate the importance of individual techniques (which is of great interests to the reader)

- the improvment from naive HMM to nested HMM does not seem to worth the effort?

- the path merging process seems to be tedious and suboptimal