---
license: mit
language:
- sah
pretty_name: spellman LID corpus — sah
size_categories:
- 100K<n<1M
task_categories:
- text-classification
tags:
- language-identification
- cyrillic
- sah
configs:
- config_name: moderated
  default: true
  data_files: data/moderated/*.parquet
- config_name: wild
  data_files: data/wild/*.parquet
- config_name: all
  data_files: data/*/*.parquet
---

# spellman LID corpus — sah

**275,761 rows / 114.0M chars** of sah-language text with
provenance, built for training and evaluating the
[spellman](https://github.com/vpermilp/spellman) language detector. Rows are
single-line texts: news articles and library works (paragraph/sentence-group
sized) plus Telegram/VK community posts and comments (wild register — emoji, hashtags,
code-switching artefacts, folk spellings kept on purpose).

Row length: median 283 chars (p10 68, p90 867).

## Configs

| config | contents | rows |
|---|---|---|
| `moderated` (default) | `news` · `library` · `wiki` — edited, published sources | 58,292 |
| `wild` | `tg-post` · `tg-comment` · `vk-post` · `vk-comment` — Telegram and VK community posts and comments | 217,469 |
| `all` | union of both | 275,761 |

```python
from datasets import load_dataset
moderated = load_dataset("vpermilp/lid-sah")            # default config
wild = load_dataset("vpermilp/lid-sah", "wild")
```

## Columns

| column | meaning |
|---|---|
| `text` | the row (single line, whitespace-flattened) |
| `lang` | ISO 639-3 (`sah`) |
| `source` | site store or Telegram channel handle |
| `register` | `news` · `library` · `wiki` · `tg-post` · `tg-comment` · `vk-post` · `vk-comment` |
| `date` | publication date when the source exposed one (YYYY-MM-DD) |
| `id` | source URL (crawl) or message/post id (Telegram, VK) |

## Registers

| register | rows |
|---|---|
| tg-post | 175,589 |
| news | 58,292 |
| tg-comment | 27,365 |
| vk-post | 14,515 |

## Sources

| source | register | rows |
|---|---|---|
| edersaas.ru-sah | news | 58,292 |
| nvk_sakha | tg-post | 34,925 |
| sakhasire20 | tg-post | 21,713 |
| sakhasire20 | tg-comment | 15,803 |
| kiin_kyorat_gazeta | tg-post | 14,982 |
| uamurusahargata | tg-post | 14,230 |
| gazetaensieli | tg-post | 12,081 |
| sanaoloh | tg-post | 11,637 |
| namskiyulus | tg-post | 11,394 |
| kyymsakha | tg-post | 11,018 |
| taattagazeta | tg-post | 9,732 |
| amma_olo5o | tg-post | 8,033 |
| suntar_sonunnara | tg-post | 7,697 |
| sakha_nvk | vk-post | 6,805 |
| sakhaa_sire | vk-post | 6,671 |
| gazetanyurba | tg-post | 6,591 |
| biikige | tg-post | 2,472 |
| uamurusahargata | tg-comment | 2,196 |
| biikige | tg-comment | 2,001 |
| valerinikolaev | tg-post | 1,550 |
| kyymsakha | tg-comment | 1,506 |
| sakhalive | tg-post | 1,485 |
| sanaoloh | tg-comment | 1,157 |
| gazetaensieli | tg-comment | 1,122 |
| archy_jiete | tg-post | 1,017 |
| dmitriytikhon0v | tg-post | 900 |
| kyymru | vk-post | 816 |
| rnk0908 | tg-post | 814 |
| mikhaleva_saya | tg-post | 796 |
| sakhasporttv | tg-post | 773 |
| namskiyulus | tg-comment | 679 |
| mekhesmasaxa | tg-post | 651 |
| gazetanyurba | tg-comment | 443 |
| sakhalive | tg-comment | 377 |
| kurulgen_uran | tg-post | 361 |
| suntar_sonunnara | tg-comment | 323 |
| rnk0908 | tg-comment | 311 |
| erelerelik | tg-comment | 305 |
| dmitriytikhon0v | tg-comment | 292 |
| sakha_video | tg-comment | 288 |
| sakha_songs | vk-post | 223 |
| sakha_video | tg-post | 215 |
| sakhasporttv | tg-comment | 208 |
| tetim_radio | tg-post | 166 |
| traditionalculturesakha | tg-post | 164 |
| traditionalculturesakha | tg-comment | 152 |
| erelerelik | tg-post | 143 |
| kurulgen_uran | tg-comment | 94 |
| tetim_radio | tg-comment | 81 |
| surukbicik | tg-post | 49 |
| kiin_kyorat_gazeta | tg-comment | 26 |
| amma_olo5o | tg-comment | 1 |

## How it was built and cleaned

1. **Orthographic gate** — every row carries letters that never occur in
   Russian orthography (e.g. Sakha ө ү ҥ ҕ һ), with a per-language
   marker-density floor; Russian-function-word share gates code-switched
   rows, which are split into native sentence runs rather than dropped.
2. **Structural cleaning** — site chrome and per-channel boilerplate
   (lines repeated across ≥8 posts), self-promo tails, TV schedules,
   Telegram thread roots (auto-forwarded post copies) removed.
3. **Structural repairs** — soft hyphens removed and print-layout splits
   rejoined (hyphenated / lowercase-continued paragraphs merged in page
   order); site breadcrumbs, TV schedules, weather tables, chord sheets,
   sponsor tails and letter-spacing artefacts dropped; Latin h/x homoglyphs
   inside Cyrillic words normalized (1,491 rows dropped);
   long rows must carry marker letters in ≥2 tokens (4,448).
4. **Content-free filter** — emoji/URL/number-only rows, Latin-dominant or
   mojibake rows, HTML-entity leftovers dropped (630).
5. **Cross-source dedup** on a normalized key (casefold; URLs, handles,
   hashtags, punctuation, digits stripped) — catches site↔channel
   cross-posts and reposts (35,310 removed).
6. **PII masking** — phone numbers and e-mails replaced by `<phone>` /
   `<email>`; no author identities were ever stored.
7. **Model judge** — rows the current spellman model contradicts at ≥0.9
   confidence outside the language's close-pair group are dropped
   (2).

No train/validation/test split is shipped: the spellman pipeline assigns
splits content-addressed (crc32) at mix time.

## License

The compilation, cleaning and annotations are released under the MIT
license. Row texts are redistributed from the public sources listed above
(newspapers, libraries, public Telegram channels and VK communities) for
language-identification research; each row carries its `source` and `id`
so upstream attribution is always recoverable.

Built 2026-08-30T12:19:59+00:00.