MANGQ: Towards Natural Language Interfaces for Knowledge GraphsAmin Harig; Sabine Janzen; Wolfgang Maaß
In: 32nd Workshop on Information Technologies and Systems. Workshop on Information Technology and Systems (WITS-2022), December 14-16, Kopenhagen, Denmark, WITS, 12/2022.
The growing number of data platforms offering large amounts of distributed, heterogeneous, but economically relevant data is more and more designed and implemented based on knowledge graphs (KGs). But despite the overall presence of KGs in data-driven systems and the growing need for intuitively accessing them, natural language querying of KGs by non-technical users is not possible. With MANGQ, we introduce a model for mapping natural language to query language expressed in GraphQL. Focusing on Question-Answering (QA) settings, we describe a question-to-query mapping approach that supports intuitive access of KGs. We present promising results from a runtime experiment with a QA system implementing the proposed approach using subsets of the CoSQL corpus and the ParaphraseBench benchmark.