---
pretty_name: Tabulus-Runs
license: other
tags:
  - scientific-literature
  - table-extraction
  - document-understanding
  - information-extraction
  - bibliography
  - citation-matching
  - scholarly-communication
  - atomic-layer-deposition
  - area-selective-deposition
viewer: false
---

# Tabulus-Runs

**Tabulus-Runs** contains recorded Atomic Layer Deposition (ALD) and
Area-Selective Deposition (ASD) experiments accompanying the Tabulus study on
reference-aware scientific table extraction.

**Software:** [https://github.com/sciknoworg/tabulus](https://github.com/sciknoworg/tabulus)  
**Documentation:** [https://tabulus.readthedocs.io/en/latest/](https://tabulus.readthedocs.io/en/latest/)  
**Gold-standard dataset:** [https://huggingface.co/datasets/SciKnowOrg/tabulus-gold](https://huggingface.co/datasets/SciKnowOrg/tabulus-gold)

The release is organized around the objects a researcher is most likely to
inspect:

**domain → paper → pipeline Step → recorded artifacts**

Tabulus-Runs contains recorded predictions and experimental outputs. The
authoritative controlled gold-standard annotations and evaluator-facing expected
outputs are published separately in **Tabulus-Gold**.

## What is Tabulus?

[Tabulus](https://github.com/sciknoworg/tabulus) is a modular framework for
digitizing scientific review tables into structured, citation-aware data. Its
seven-Step workflow covers table localization and reconstruction, identification
of reference-bearing tables, bibliography extraction, citation-to-bibliography
matching, scholarly-reference resolution, and export of resolved structured
tables.

Tabulus-Runs records artifacts produced while applying these Steps to review
papers from the ALD and ASD literature.

## Dataset structure

```text
tabulus-runs/
├── README.md
├── ALD/
│   ├── README.md
│   ├── Cremers_2019/
│   ├── Miikkulainen_2013/
│   └── Puurunen_2005/
├── ASD/
│   ├── README.md
│   ├── Bent_2025/
│   ├── Bonvalot_Vallee_2021/
│   ├── Choi_2023/
│   ├── Mackus_2014/
│   ├── Park_2026/
│   ├── Parsons_2020/
│   └── Tonner-Zech_2025/
└── analysis/
    ├── README.md
    ├── all_runs.csv
    ├── step5_runs.csv
    ├── step5_ambiguous.csv
    ├── step5_skipped.csv
    └── step5_unmatched.csv
```

The release covers **10 source papers**: **3 ALD papers** and **7 ASD papers**.
Each paper directory contains the recorded artifacts retained for the seven
pipeline Steps and, where applicable, machine-actionable examples.

The source publication PDFs are **not distributed** as part of the public
dataset. Paper-level documentation identifies the corresponding source
publication.

## Recorded pipeline artifacts

| Step | Recorded task | Public artifact role |
|---|---|---|
| **Step 1 — Table localization** | Locate and profile table regions in the source document | Recorded localization, crop, and related table-detection artifacts |
| **Step 2 — Table reconstruction** | Reconstruct detected tables as structured tabular data | Adapter-specific reconstructed table outputs |
| **Step 3 — Reference-table classification** | Identify reconstructed tables containing explicit literature-reference information | Recorded classification outputs exposed under `step3/` |
| **Step 4 — Bibliography extraction** | Extract the ordered bibliography of the source publication | Recorded bibliography artifacts |
| **Step 5 — Reference matching** | Match citations in table cells to bibliography positions | Recorded citation-to-bibliography matching outputs exposed under `step5/` |
| **Step 6 — Scholarly reference resolution** | Resolve linked bibliography entries to scholarly identities | Recorded paper-level resolution outputs |
| **Step 7 — Resolved-table export** | Join table content, bibliography links, and resolved scholarly identities | Deterministically generated resolved-table outputs |

Step 3 and Step 5 artifacts were originally produced within adapter-specific
reconstruction runs. They are exposed as explicit `step3/` and `step5/`
directories in this public release for clarity.

Public directory and filename terminology uses **Step**, not **Stage**.
Historical field names inside preserved machine-generated metadata are retained
where changing them would alter the recorded artifact.

## Release contents

The public release contains:

- **10** paper-level experiment directories
- **153** Step 3 classification artifacts
- **153** Step 5 reference-matching artifacts
- **10** Step 6 paper-level reference-resolution artifacts
- **153** Step 7 manifests
- **1,327** Step 7 resolved CSV outputs

The `analysis/` directory additionally contains experiment-level summaries and
Step 5 diagnostic tables retained for research inspection.

These counts describe recorded artifacts rather than independent gold-standard
evaluation instances. Controlled quality targets are provided separately in
**Tabulus-Gold**.

## Recorded-run provenance

Operational cluster logs and `_run_metadata/` directories are intentionally
excluded from the public release. Research-relevant run summaries are retained
under `analysis/`.

Legacy absolute filesystem paths embedded in retained artifacts were normalized
to portable provenance identifiers such as `original-run://...`. These
identifiers preserve the logical provenance relationship of the recorded
artifact but are not expected to resolve on another machine.

The public release preserves scientific experiment outputs while excluding
machine-specific operational information that is not required to inspect or
reuse the recorded results.

## Relationship to Tabulus-Gold

[**Tabulus-Gold**](https://huggingface.co/datasets/SciKnowOrg/tabulus-gold)
contains the authoritative annotations and expected outputs used for controlled
quality evaluation of the Tabulus pipeline.

**Tabulus-Runs** contains recorded experimental artifacts from the wider ALD and
ASD study corpus.

Keeping the resources separate distinguishes the **evaluation targets** from the
**system outputs and recorded experiments** being evaluated.

## Citation

When using **Tabulus-Runs**, please cite the dataset repository. A DOI-based
citation may be added if an archival DOI record is published in the future.

If you use the **Tabulus software**, please also cite the software repository:

```bibtex
@software{tabulus,
  author = {D'Souza, Jennifer and Rumleanschi, Vladimir},
  title = {Tabulus: End-to-End Scientific Review Table Digitization Pipeline},
  url = {https://github.com/sciknoworg/tabulus},
  year = {2026}
}
```

Tabulus was first developed in the following master's thesis:

```bibtex
@mastersthesis{rumleanschi2026tabulus,
  author = {Rumleanschi, Vladimir},
  title = {Tabulus: AI-Assisted Tabular Knowledge Extraction and Semantic Integration Framework for Scientific Literature},
  year = {2026},
  doi = {10.15488/21618},
  url = {https://doi.org/10.15488/21618}
}
```

## License

The original experiment metadata, recorded Tabulus outputs, analysis files, and
other original content created for **Tabulus-Runs** are licensed under the
[Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0)](https://creativecommons.org/licenses/by-sa/4.0/).

You are free to share and adapt these materials for any purpose, provided that
appropriate credit is given to the creators and that adapted material is
distributed under the same or a compatible license.

Third-party material derived from the source publications is not covered by this
license unless explicitly stated otherwise. Copyright and reuse rights for such
material remain with the respective copyright holders.