The Chinese Multi-Word Expression Extraction Based on Improved Semi-supervised Algorithm
Liang Yinghong, Hong Lei, Yannan Liu · 2017
Failing to identify multi-word expression (MWE) may cause serious problems for many Natural Language Processing (NLP) tasks. Because of lack of Chinese MWE labeled corpus, we used a semi-supervised Co-trade algorithm to extract Chinese MWE, furthermore, a supervised head word expansion method was used to find cluster center in data editing step, which makes that the supervised information was also added to the Co-trade algorithm in order to select correct data. Experiment result shows that the extraction results of Chinese multi-word expression using improved Co-trade algorithm is better than that of original algorithm, which verified that the improved algorithm is effective.