This repository contains code for a Chinese open-domain knowledge base question answering system that placed fourth in the CCKS2019 CKBQA competition. The solution references the second-place approach from CCKS2018 COQA and extends it with BERT-based sequence labeling for named entity recognition, semantic matching models, and entity linking over the PKUBASE knowledge graph. Users need to set up a Neo4j graph database locally to run entity extraction and complex question answering pipelines including bridging steps.