Rbosaghz review of Poon 2007 joint inference in information extraction

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

This paper deals with extracting fields from the citation text available at the end of papers.

When extracting author names, conferences/journal names, year of publication, etc, most current systems work by first looking at a potential record's text, and extracting from it the relevant fields. Then once the fields have been extracted, some records may denote the same paper, and so we would like them to be merged together. Current systems do this step separately from segmenting the record text into entities. There is an obvious advantage to doing the extraction and merging steps together, and that's what this paper focuses on. They use Markov logic to solve the entire task at once.

I liked this paper because it is the first application of Markov Logic that I have seen.