A New Word Discovery Method Based on Constrainted and DSG

Accurate identification of risk places is of great significance for risk identification, risk analysis, risk response planning and other work in risk areas. Therefore, we will combine the word formation features of risk place and statistical algorithms to discover new words in this paper. Firstly, we preprocess the data based on conditional random field (CRF), and develop a professional dictionary with the help of expert. Secondly, segment the test set, and form a set of new words to be selected based on segmentations and their left-right words. Finally, we make word vectors based on directional skip-gram (DSG), and calculate the distance between two words to discover new words. The results show that the algorithm can accurately identify new places, and it is suitable for discovering professional vocabulary which word-formation features is compound.