JOINT INFERENCE IN INFORMATION EXTRACTION USING MARKOV LOGIC
J. Refonaa · 2013
Information extraction takes text or semi-structured data as the input and produces structured records as the output.Markov Logic is a Statistical Relational Learning technique used for learning from data which has relational struture. The goal of the project is to study the performance of a Statistical Learning Technique (SRL) such as Markov Logic for Information Extraction tasks. We use citation matching as a testbed for illustrating the use of Markov Logic for information extraction. Citation matching involves extracting bibliographic records from citation lists in technical papers and merging records that represent the same publication. We focus only on extracting titles, authors and venues from citation strings. We extract these fields, first by first segmenting each candidate record separately, and then merging records that refer to the same entities. A joint approach to information extraction is used where segmentation of all records and entity resolution are performed together in a single integrated inference process. We use Alchemy, a open source tool, for uaing Markov Logic and then intend to incorporate a different learning/inference algorithms for information extraction.