Semi-automatic construction of a named entity dictionary based on active learning

초록

A named entity (NE) dictionary is an important resource to affect the performance of NE recognition, but it is not easy to construct the NE dictionary manually, because human annotation is time-consuming and labor-intensive. We propose a semi-automatic model to construct an NE dictionary from the free online resource DBpedia. The proposed model expands and purifies an NE dictionary based on an active learning technique. In the experiments, the proposed model classified 99.99% (180,008 out of 180,020 entries) of DBpedia entries into 18 NE categories with macro-averaging F1-measures of 0.6980 for 18 NE categories (0.7519 for 17 NE categories). © Springer-Verlag Berlin Heidelberg 2015.

키워드