Fig. 8

Integration of named entity recognition into active learning with PAS and n-grams against GRO’13 dataset. The learning curves for the TEES system under the proposed informativity method using predicate-argument relation (PAS) and unigram model, as well as the conventional committee (ConventionalCommittee) based active learning method as the benchmark. In contrast, each method is integrated with the output from the named entity recognition result (NE)