---
language:
- he
- arc
- en
license: cc-by-4.0
task_categories:
- sentence-similarity
- text-retrieval
tags:
- rabbinic
- hebrew
- aramaic
- talmud
- cross-lingual
- bitext
- sefaria
size_categories:
- 1K<n<10K
---

# Rabbinic Hebrew/Aramaic - English Parallel Corpus

A benchmark dataset for evaluating embedding models on Rabbinic Hebrew and Aramaic texts, with parallel English translations sourced from [Sefaria](https://www.sefaria.org).

## Dataset Description

This dataset contains 3,708 parallel text pairs spanning diverse Rabbinic literature across multiple centuries and genres. It is designed for evaluating cross-lingual embedding models on their ability to align Hebrew/Aramaic source texts with English translations.

### Languages

- **Source**: Rabbinic Hebrew, Jewish Babylonian Aramaic, Jewish Palestinian Aramaic
- **Target**: English

### Dataset Structure

Each example contains:
- `ref`: Sefaria reference string (e.g., "Berakhot.2a:1")
- `he`: Hebrew/Aramaic source text
- `en`: English translation
- `category`: Text category

### Categories

| Category | Count | Description |
|----------|-------|-------------|
| Mishnah | 789 | Tannaitic legal compilation (~200 CE) |
| Tanakh Commentary | 674 | Rashi, Ramban, Radak, Rabbeinu Behaye on Torah |
| Jerusalem Talmud | 520 | Palestinian Talmud (~400 CE) |
| Talmud | 480 | Babylonian Talmud (~500 CE) |
| Midrash Rabbah | 393 | Midrashic compilations |
| Hasidic/Kabbalistic | 304 | Likutei Moharan, Tomer Devorah, Kalach Pitchei Chokhmah |
| Philosophy | 240 | Guide for the Perplexed, Sefer HaIkkarim |
| Halacha | 160 | Sefer HaChinukh, Mishneh Torah |
| Mussar/Ethics | 108 | Chafetz Chaim, Kav HaYashar, Iggeret HaRamban |
| Targum | 40 | Aramaic Targum to Song of Songs |

## Intended Use

### Primary Use Case

Evaluating embedding models for cross-lingual retrieval:
- Given a Hebrew/Aramaic text, can the model find its English translation from a pool of candidates?
- Models that excel at this task likely capture the semantics of Rabbinic literature well.

### Evaluation Metrics

- **Recall@k**: Percentage of queries where correct translation is in top k results
- **MRR**: Mean Reciprocal Rank
- **Bitext Accuracy**: True pair vs random pair classification

## Source

All texts and translations are from [Sefaria](https://www.sefaria.org), a free library of Jewish texts.

### Translations

Translations come from various sources including:
- William Davidson Talmud (Steinsaltz)
- Sefaria Community translations
- Historical translations (e.g., Friedlander's Guide for the Perplexed)

## Citation

If you use this dataset, please cite Sefaria:

```bibtex
@misc{sefaria,
  title = {Sefaria: A Living Library of Jewish Texts},
  url = {https://www.sefaria.org},
  year = {2024}
}
```

## License

The dataset is released under CC-BY 4.0, following Sefaria's licensing for their open texts.