A Study on Automatic Extraction of New Terms

This research explores to automatically predict new terms based on linguistic features and statistical behaviors of noun phrases during a special period. It integrates both syntactic function value and TF-IDF value into an automatic term extraction system to weight new term candidates. Research questions include: what are the linguistic and statistic properties of new terms during a special period? Will linguistic features contribute to prediction of new terms? And will statistic features like, TFIDF Value contribute to prediction of new terms? Correspondingly, a series of experiments are conducted on medical corpus to examine a group of new terms' distribution properties and syntactic features across two years in comparison. The results show there does exist significant difference between two groups of values. Regardless of this limitation, this research is meaningful as it attempts to realize automation of selection process of new medical terms, which will greatly avoid subjective decisions and reduce experts' workloads.