A Prototype Patterns Selection Algorithm Based on Semi-supervised Learning

Semantic role labeling (SRL) is a fundamental task in natural language processing to find a sentence-level semantic representation. At present, the mainstream studies of semantic role labeling focus on the use of a variety of statistical machine learning techniques. But it difficult to obtain high quality labeled data. To solve the problem, we proposed a novel prototype patterns selection algorithm based on semi-supervised learning in this paper. There are two main innovations in this article: firstly, order parameter evolution is introduced to expand training data. The strongest order parameter will win by competition and desired pattern will be selected. Secondly, the must-links and cannot-links constraints exist in the train data is used to reduce the noise of extend data. The experiment results show the proposed method has a higher performance for semantic role labeling.