MatMech: A Multimodal Dataset for Exploring Casual Mechanisms in Materials Science Literature

OverviewThe <b>MatMech Dataset</b> is a large-scale collection of <i>59,200+</i> materials science papers, each represented by a structured JSON file and associated figure images.This dataset provides both the <b>full raw collection (compressed)</b> and <b>sample cases </b>for quick exploration.Data CompositionThe dataset contains:matmech.zip: Full dataset (\~19 GB) containing all papers, organized by DOI folders. Each folder includes:<br>• <i>paper.json</i>: parsed paper metadata and text content<br>• <i>images/</i>: figures extracted from the PDFcase/: A curated set of small example cases for quick testing and inspection dataset_summary.json: Machine-readable dataset metadataREADME.md: Human-readable dataset description and usage instructionsThe directory structure is shown as follow: matshare_dataset/<br>│<br>├── README.md<br>├── matshare_full.zip<br>├── case/<br>│ ├── doi_001/<br>│ ├── doi_002/<br>│ └── ...<br>└── data_summary.json<br><br><b><i>For more information, please refer to the </i></b><i>README.md</i>