RAGonite is a flexible RAG system built by the NLP team at Fraunhofer IIS. It is a conversational question answering (ConvQA) system that can:
- operate over heterogeneous data elements like documents with passages, lists, and tables, and RDF knowledge graphs;
- explain its answers via counterfactual attribution;
- contextualize evidences in collections using context from original documents;
- perform iterative and hybrid (SQL, lexical, dense) retrieval over a mixture of input sources, with subsequent answer fusion;
- automatically induce databases from knowledge graphs (KGs); and
- run over multiple domains like enterprise wikis, automobile KGs, soccer, and fictional universes.
Specifically, this page contains all open artifacts (data, code, screenshots, videos, slides, posters, ...) arising from the following publication(s):
-
Rishiraj Saha Roy, Joel Schlotthauer, Chris Hinze, Andreas Foltyn, Luzian Hahn and Fabian Küch, "Evidence Contextualization and Counterfactual Attribution for Conversational QA over Heterogeneous Data with RAG Systems." in Proceedings of the 18th ACM International Conference on Web Search and Data Mining 2025 (WSDM '25) , 10 -14 March 2025, Hannover, Germany.
-
Rishiraj Saha Roy, Chris Hinze, Joel Schlotthauer, Farzad Naderi, Viktor Hangya, Andreas Foltyn, Luzian Hahn and Fabian Küch, "RAGONITE: Iterative Retrieval on Induced Databases and Verbalized RDF for Conversational QA over KGs with RAG." in Proceedings of the 21st Conference on Database Systems for Business, Technology and Web (BTW '25) , 3 -7 March 2025, Bamberg, Germany.
These supplementary materials can be found in the respective directories (see the "wsdm25-confluence" folder for the WSDM 2025 paper's artifacts and "btw25-bmw" folder for the BTW 2025 paper's artifacts).