Method and system for generating patent abstract based on multi-task feature collaboration

The invention discloses a method and a system for generating patent abstracts based on multi-task feature collaboration, which are used for respectively generating technical problems to be solved and generating texts such as invention information in beneficial effects, purposes, core schemes, invention points and other technical schemes based on a multi-task model. The final patent abstract is obtained through the process of summarizing and generating the patent abstract, the purpose of automatically writing and outputting the patent abstract by inputting the original patent text is achieved, the labor cost of patent abstract writing is reduced, and the patent abstract writing efficiency is improved. 本发明公开了一种基于多任务特征协同的生成专利摘要的方法及系统，基于多任务模型分别生成要解决的技术问题和有益效果、用途、核心方案、发明点、其他技术方案中的发明信息等生成文本，通过汇总生成专利摘要流程得到最终的专利摘要，实现输入原始专利文本自动撰写输出专利摘要的目的，降低了专利摘要撰写的人工成本，提升了专利摘要撰写的效率。