Zhengyuan Dong

Current approaches for Natural Language to SPARQL (NL2SPARQL) generation primarily rely on one-turn, training-intensive models. While effective in specific settings, these models often lack generalizability and fail to provide transparency or mechanisms for error recovery in realistic scenarios....