An improved algorithm for relation extraction based on tri-training

The tri-training algorithm is an efficient co-training method for semi-supervised learning, and it has been used to extract semantic relation between entities in text. However, the tri-training method will introduce noises and lose some valuable samples while expanding the training set. In this paper, we propose an improved algorithm based on tri-training. New voting mechanism and active learning method are introduced into the improved algorithm to solve the problems of traditional tri-training algorithm. Experiments demonstrate that the performance of the improved algorithm is superior to the existing tri-training algorithms.