Design And Implementation Of Medical Question And Answer Platform Based On Knowledge Graph

The improving national industry in China has raised people’s standard of life,but it has also increased the number of physical sicknesses.Nevertheless,there are many circumstances in China where it is tough to see a doctor and register.In order to solve this situation effectively,we have constructed a medical Q&A platform based on knowledge graph.To solve the problem that it is difficult for the people to visit the doctor.Traditional search engines provide a large volume of search results through keyword search.Although they can solve healthcare problems,their precision is very low.In recent years,knowledge graph has grown swiftly and has been used in many fields.It portrays entities,properties and connections through graphs to form a specific diagram of the web and to build an active and effective knowledge system.The knowledge graph is used as the knowledge base of medical Q&A system to finish the building of medical Q&A system and provide the accurate answers for medical customers.This paper focuses on how to realize the building of the KM graph and the question answering system,and how to enhance the question answering accuracy as much as possible.The main components of the work are as follows.1.data capture and pretreatment: using Python reptiles scripts to capture the required messages from two medical portals "xunyiwenyao" and "xiuyiwang",and storing the data in pre-set property fields in the Mysql database and then performing xpath analysis on the saved data to export Json data,and then data are washed and reprocessed manually,and finally the more normalized Json data are output.2.2.Knowledge map construction.The system employs bottom-up design approach in KM construction,where entity,property and relative information are collected from Json data,and the acquired information is used to structure the KM map,and the information is saved to Neo4 j while the content is in the DB.It is easier to view the mapping of health care knowledge and conducive to the follow-up content query and analysis.3.the construction of Q&A system process.This paper investigates the structure of medical Q&A system based on the integration of semiconductor analysis and question pattern matching.The Q&A system research is classified into four main modules: question sorting,named entity identifying,template pairing and matching,and querying.The question classification module mainly uses Python’s AC algorithm to achieve the initial classification of questions;the entity recognition module mainly uses Bi-LSTM+CRF network model for named entity identification,and of course uses word division techniques for term separation before entity identification;the template matching module mainly combines the input questions with the results of question classification and entity recognition,and then with the question module library Match,and finally The matched templates are integrated with the analyzed entity and question classification results and translated into Neo4 j query sentences for knowledge graph query.