ElecGraph-RAG: A Star-Shaped Graph Modeling and Efficient Hierarchical Retrieval Framework for Electronic Datasheets

ACL ARR 2026 January Submission9347 Authors

Keywords: Information Retrieval, Retrieval-Augmented Generation (RAG), Knowledge Graph, Electronic Design Automation (EDA), Document AI

Abstract: Existing graph-based retrieval-augmented generation (RAG) methods often fail to capture the device-centric, star-shaped ontology of electronic datasheets, leading to topological mismatches and entity--attribute misattribution in technical question answering. We propose ElecGraph-RAG, an ontology-aligned hierarchical RAG framework explicitly designed for device-centric knowledge organization, which introduces ontology-aware semantic abstraction as an inductive bias for retrieval. The framework performs coarse-to-fine retrieval by first localizing relevant high-level semantic abstractions and then extracting fine-grained attribute-level evidence. We also introduce ElecGraph-QA, a benchmark of 200 complex factual and comparative queries derived from real-world industrial specifications. Experimental results show that ElecGraph-RAG achieves 94.28% and 85.00% accuracy on factual and comparative queries, respectively, while reducing token consumption by 76.3% compared to state-of-the-art graph-based baselines. These results demonstrate the effectiveness of ontology-aligned hierarchical abstraction for efficient domain-specific RAG.

Paper Type: Long

Research Area: Information Extraction and Retrieval

Research Area Keywords: Information Extraction, Information Retrieval and Text Mining, Machine Learning for NLP, Question Answering

Contribution Types: NLP engineering experiment, Approaches low compute settings-efficiency

Languages Studied: English

Submission Number: 9347