Knowledge Graph-based Q&A System For College Admission Information Research And Implementation

Automated question answering research is a branch of natural language processing and one of the hot topics in the field.From the earliest Turing test to today’s deep learning models,automated question answering has been continuously improving.Knowledge graph is a knowledge network formed by transforming knowledge into triple form and stored in a graph database,avoiding the complexity of relational database calls.An automated question answering system in a specific field based on the knowledge graph can traverse entities and entity relationships in the knowledge graph according to the user’s natural language query and answer the user’s questions through certain methods.Nowadays,automated question answering systems based on knowledge graphs have gradually developed,with research hotspots mainly focusing on areas such as medicine and agriculture,but relatively few in the education field.This article aims to build a knowledge graph automated question answering system for college admission information to serve the admission consultation.The main research work of this thesis includes:1.Collecting multi-source data based on university enrollment information through web crawler technology,cleaning and annotating the data to form the required dataset for the knowledge graph of enrollment consultation services.2.Building multiple named entity recognition models for experimentation and comparison on different datasets,and comparing different models on the same dataset.Then comparing the pipeline model with the joint extraction model to select the best model for constructing the knowledge graph.3.Analyzing the problem set dataset to determine problem intent categories,recognizing intent through the use of the ALBERT-TEXTCNN model.Designing slot filling templates to fill the slots.4.Constructing a front-end and back-end separated automatic question answering system,with the front-end built using Vue and Java Script,and the back-end using Python.Simultaneously,to ensure the real-time and accuracy of the knowledge graph,a knowledge graph management system has been constructed.The innovation of this study lies in(1)proposing a new named entity recognition model combining ALBERT and CRF,which is experimentally compared with other models and shows superior performance;(2)constructing a knowledge graph management system after the question-answering system to ensure information accuracy,real-time updating,and maintainability.