README.md

---
language:
- si
tags:
- ocr
- sinhala
- diachronic
pretty_name: "SiDiaC: Sinhala Diachronic Corpus"
arxiv: 2509.17912
---

# SiDiaC: Sinhala Diachronic Corpus

A curated collection of Sinhala literary texts and related resources with OCR outputs and basic metadata for each work. This repository organises original PDFs alongside finalised OCR text and machine-readable `metadata.json` files per title.

Paper: [https://arxiv.org/abs/2509.17912](https://arxiv.org/abs/2509.17912)

**GitHub Repository:** [https://github.com/NeviduJ/SiDiaC](https://github.com/NeviduJ/SiDiaC)


### Example metadata.json
```json
{
  "title": "පන්සිය පනස් ජාතක පොත",
  "title_en": "Pansiya Panas Jathaka Potha",
  "author": "Unknown",
  "author_en": "Unknown",
  "genre": "Fiction; Religious",
  "issued_date": "1881",
  "written_date": "1303 - 1333",
  "ocr_confidence": 0.9987
}
```

### Usage

You can load this dataset using the `datasets` library:

```python
from datasets import load_dataset

ds = load_dataset("Nevidu/SiDiaC", split="train")

# Verify the number of rows
print(f"Loaded dataset with {len(ds)} rows")

# Print the first example
print(ds[0])
```

### Citation

```json
@article{jayatilleke2025sidiac,
  title={SiDiaC: Sinhala Diachronic Corpus},
  author={Jayatilleke, Nevidu and de Silva, Nisansa},
  journal={arXiv preprint arXiv:2509.17912},
  year={2025}
}
```