MEDLedge: A QA based system for constructing medical knowledge base

摘要

Most existing works have focused on constructing a knowledge base use machine learning techniques. However, they often suffer from obtaining incorrect entity relationships. In this paper, we propose a system, called MEDLedge, to establish a reliable medical knowledge base. The aim is to improve the accuracy and comprehensiveness of medical knowledge base. The system consists of three parts: data processing, data analytics and expert Q&A platform. In data processing phase, we extract entities and relationships from clinical data using a hierarchical segmentation method. Then, the results of the extraction that can be found directly in a medical dictionary are stored in knowledge base. For the others, we design the crowdsourcing questions and submit them to the expert Q&A system. We use majority vote algorithm to select correct relationships from the answers and store them into our knowledge base. We implement a system to construct the knowledge base, and will release it in the future.