GRAFT: Graph-Matched Retrieval and Fusion of Tables in Data Lakes

2026-07-14

AI Summary

This work addresses the limitation of existing table retrieval methods, which often overlook inter-table joinability and combinability, leading to fragmented evidence that hinders downstream tasks. The authors formulate table retrieval as a graph matching problem between a query intent graph and a heterogeneous data lake graph, proposing a three-stage online pipeline to efficiently generate fused subgraphs. Key innovations include the IGMS scoring function, which unifies semantic relevance, structural compatibility, and evidence diversity, and a formalization of subgraph generation as a Markov decision process, enabling efficient search through implicit Q-learning combined with a canonical compression operator. Evaluated on the Spider and BIRD benchmarks, the approach achieves a 7.8% gain in F1 and a 10.6% improvement in Sufficiency over state-of-the-art methods while maintaining high retrieval efficiency.

Abstract

Autonomous data agents resolve analytical queries by retrieving and reasoning over evidence in tabular data lakes. Existing methods score tables independently against the query and ignore the joinability and unionability that link them, returning fragmented evidence that downstream agents cannot integrate. We propose GRAFT (Graph-matched Retrieval and Fusion of Tables), structured around two principal contributions. First, we cast table retrieval as a graph matching problem between a query-derived intent graph and a heterogeneous data lake graph, and introduce IGMS, a log-determinant reward that couples semantic relevance, structural compatibility, and evidence diversity in a single objective. Second, we recast subgraph generation as a Markov decision process and learn a value function via implicit Q-learning on self-generated trajectories produced by a canonical compression operator that inverts the homomorphism. We further design a three-stage online pipeline that exploits anchor reachability, predicate admissibility, and reward monotonicity to greatly prune the candidate space before exact IGMS evaluation. On Spider and BIRD adapted to the tabular data lake setting, GRAFT achieves the best Recall, Precision, F1, and Sufficiency among point-wise, greedy-expansion, and structure-aware baselines, with relative gains of 7.8% in F1 and 10.6% in Sufficiency over the strongest baseline, while maintaining high search efficiency.

Problem

• table retrieval
• data lakes
• joinability
• unionability
• evidence integration

Innovation

• graph matching
• table retrieval
• data lake
• implicit Q-learning
• IGMS