QRKB-16k (Synthetic Query Reformulation for Knowledge Graphs) is a dataset of 16,384 synthetic query reformulation pairs designed to facilitate research in query understanding and retrieval-augmented generation (RAG) using knowledge graphs.
Each pair consists of a natural language query and a set of corresponding subqueries, with each subquery structured as a partial semantic triple, suitable for retrieval from knowledge graphs like DBpedia and… See the full description on the dataset page:
https://huggingface.co/datasets/alexdong/query-reformulation.
