中文专利关键信息语料库的构建研究

As a kind of important technology document, the patent is of substantial significance to the national intellectual property strategy in China. Existing patent corpus are mostly for the purpose of information retrieval and machine translation task, leaving the fine-grained annotated patent less touched. To facilitate the forth-coming intelligent patent technology development, this paper constructs a Patent Key Information Corpus, consisting of 313 patents annotated with the issues, methods and effects in the texts. Then the SOTA named entity recognition models are applied to the corpus, and the sharping decrease in the performance indicate the automatic identification of the key information in a patent is a challenging IE task.