Named entity recognition (NER) is a major subtask of information extraction. Previous research tent to use huge amount of labeled data to train a classifier. But it is expensive for low resource languages One of the dominant problems facing Tibetan named entity recognition is the lack of training data. Active learning is a supervised machine learning algorithm which can achieve greater accuracy with fewer training labels. Active learning has been successfully applied to a number of natural language processing tasks, such as, information extraction, named entity recognition, text categorization, part-of-speech tagging, parsing, and word sense disambiguation. In this paper, we apply active learning based on Conditional Random Field (CRF) for Tibetan named entity recognition to minimize labeling effort by selecting the most informative instances to label. This paper proposes two kinds of query strategies, including Confidence, and Named Entity features. We compare the query strategies with the random method, and show that considerable performance improvements in reduce the human effort.
@InProceedings{LIU18.2, author = {Fei-Fei Liu and Zhi-Juan Wang}, title = {Active Learning for Tibetan Named Entity Recognition based on CRF}, booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)}, year = {2018}, month = {may}, date = {7-12}, location = {Miyazaki, Japan}, editor = {Jinhua Du and Mihael Arcan and Qun Liu and Hitoshi Isahara}, publisher = {European Language Resources Association (ELRA)}, address = {Paris, France}, isbn = {979-10-95546-15-3}, language = {english} }
Summary of the paper