Automatic Information Extraction in Semi-structured Official Journals

摘要

Information extraction systems are used to extract only relevant text information in digital repositories. The current work proposes an automatic system to extract information in semi-structured official journals. In our approach, given an input document, a Machine Learning (ML) algorithm classifies the document’s fragments into class labels which correspond to the data fields to be extracted. The implemented system deployed different features sets and algorithms used in the classification of the fragments. The system was evaluated through experiments on a sample containing 22770 lines of the Pernambuco’s Official Journal. The experiments performed revealed, in general, good results in terms of precision, which ranged from 70.14% to 98.63% depending on the feature set and algorithm used in the classification of the fragments.