基于知识图谱的智能问答系统

pip install -e .

export LLM="gpt-4o"                 # "gpt-4-turbo" as an alternativeexport SEMANTIC_SCHOLAR_KEY=""      # fill API key for semantic scholarexport OPENAI_API_KEY=""            # fill opanAI api keyexport SYNONYM_API_KEY=""           # Bayer internal linnaeusannotate synonym API keyexport SYNONYM_API_URL=""           # Bayer internal linnaeusannotate synonym API url

streamlit run src/fact_finder/app.py --browser.serverAddress localhost

streamlit run src/fact_finder/app.py --browser.serverAddress localhost -- [args]streamlit run src/fact_finder/app.py --browser.serverAddress localhost -- --normalized_graph --use_entity_detection_preprocessing

--normalized_graph  =  Apply synonym replacement based on the normalized graph to the cypher queries before applying them to the graph.--use_entity_detection_preprocessing  =  Apply entity detection to the user question before generating the cypher query. The found entities will be replaced by their preferred terms and a string describing their category (e.g. "Psoriasis is a disease.") will be added to the query. This requires the corresponding api key ($SYNONYM_API_KEY) to be set. Also, the normalized graph should be used.--use_subgraph_expansion  =  The evidence graph gets expanded through the surrounding neighborhoods.