As a result of historical reasons and writing habits, the effects of medicine in Traditional Chinese Medicine (TCM) patents are often described using four character phrases. These four character phrases are not easily identified by the Chinese word segmentation system, thus greatly affects the results of patent analysis and mining. This paper proposes a semi-supervised learning method to collect four character effect phrases from the abstracts texts of TCM patents, which can help enrich the lexicon of Chinese word segmentation system, and also provide support for semantic patent retrieval and analysis. The experimental results show the validity of the method.
展开▼
