File size: 1,355 Bytes
14fa71b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0d3cf3f
514d128
0d3cf3f
514d128
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
---
license: unknown
task_categories:
  - text-classification
language:
  - bn
pretty_name: Bethik lexicon (Bangla words + grapheme map)
tags:
  - bengali
  - bangla
  - spell-check
  - grammatical-error-detection
  - bethik
  - bhashabhrom
---

# bethik-lexicon

Bangla word list + grapheme→Latin map used by the **Bhashabhrom / bethik** GED pipeline.

| File | Role |
| --- | --- |
| `words.csv` | Source lexicon (~667 MiB). Column `words`. |
| `gmap.json` | Grapheme → Latin transliteration map for `create_resources.py`. |
| `SHA256SUMS` | Content hashes of the two data files. |

## Regenerate derived artifacts (not hosted here)

From private GitHub [`BengaliAI/bethik`](https://github.com/BengaliAI/bethik) `MODELING/create_resources.py`:

```text
words.csv + gmap.json  →  trans.csv, trans.json, weights.npy
```

Those three (~5.9 GB) are regenerable — do not treat Drive copies as archive of record.

## Related

- Competition front door: [Kaggle bengali-ged](https://www.kaggle.com/competitions/bengali-ged)
- Code + small folds / annotations / private test: private repo `BengaliAI/bethik`

## License

`unknown` — not cleared for redistribution assumptions beyond current public Hub listing. Rights review pending.

## Integrity

`words.csv` / `gmap.json` SHA256 roundtrip verified 2026-07-17 against upload manifest (`SHA256SUMS`).