| --- |
| tags: |
| - sea-bed |
| - deduplicated |
| dataset_info: |
| features: |
| - name: source |
| dtype: large_string |
| - name: target |
| dtype: large_string |
| - name: source_row_index |
| dtype: int64 |
| - name: flags_fired |
| dtype: large_string |
| splits: |
| - name: train |
| num_bytes: 297204 |
| num_examples: 1805 |
| download_size: 134899 |
| dataset_size: 297204 |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: data/train-* |
| --- |
| |
| # software-documentation-ind-bitextmining |
|
|
| Deduplicated copy of [`kornwtp/software-documentation-ind-bitextmining`](https://huggingface.co/datasets/kornwtp/software-documentation-ind-bitextmining), |
| part of the SEA-BED data-quality work. |
|
|
| - **Source dataset:** `kornwtp/software-documentation-ind-bitextmining` |
| - **Deduplicated on:** 2026-09-04 |
| - **Task type:** bitext_mining |
| - **Splits:** train |
| |
| ## What changed |
| |
| Duplicate (source, target) pairs collapsed to one row, and per-side duplicate sources/targets collapsed so the split has a retrieval ceiling of 1.0 on both sides. Every removed row was an exact duplicate after normalization; no row was removed for being low quality. |
| |
| Row counts and the exact policy used are recorded in `bitext_mining_dedup_summary.csv` in the |
| analysis repo. This dataset is *not* identical to the published SEA-BED data -- |
| scores computed on it are not directly comparable to published SEA-BED numbers. |
|
|
