Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,77 @@
|
|
1 |
-
---
|
2 |
-
license: cc-by-nc-4.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-nc-4.0
|
3 |
+
language:
|
4 |
+
- ru
|
5 |
+
base_model:
|
6 |
+
- ai-forever/FRED-T5-1.7B
|
7 |
+
---
|
8 |
+
|
9 |
+
# Grammatical Error Detection for Russian
|
10 |
+
|
11 |
+
This model detects grammatical errors, misspellings, and typos in Russian text, as detailed in the preprint [arXiv:2505.04507v1](https://arxiv.org/abs/2505.04507).
|
12 |
+
|
13 |
+
## Model Functionality
|
14 |
+
|
15 |
+
- **Task:** Binary classification for the presence of defects in text.
|
16 |
+
- **Input:** Russian text, which can range from a single sentence to a full paragraph.
|
17 |
+
- **Output:** `True` if any errors are detected; `False` if the text is error-free.
|
18 |
+
|
19 |
+
## Key Notes
|
20 |
+
|
21 |
+
- The model is designed to handle multi-sentence context.
|
22 |
+
- It only performs detection and provides a binary output. It does not locate the specific errors or suggest corrections.
|
23 |
+
|
24 |
+
## Usage example
|
25 |
+
|
26 |
+
```python
|
27 |
+
import torch
|
28 |
+
import transformers
|
29 |
+
|
30 |
+
ged_model_path = "inkoziev/ged-FRED-T5-1.7B"
|
31 |
+
ged_tokenizer = transformers.AutoTokenizer.from_pretrained(ged_model_path)
|
32 |
+
ged_model = transformers.T5ForConditionalGeneration.from_pretrained(ged_model_path, device_map='cuda:0', torch_dtype=torch.half)
|
33 |
+
|
34 |
+
# Input texts to check against grammatical or orthographical defects.
|
35 |
+
input_texts = ["Расскажу как настроить плагин и сделать быструю домашнюю страничку, расскажу как найти продвинутые гайды для создания сверх-эстетичной домашней страничке.",
|
36 |
+
"Расскажу, как настроить плагин и сделать быструю домашнюю страничку, расскажу, как найти продвинутые гайды для создания сверхэстетичной домашней странички."]
|
37 |
+
|
38 |
+
# Construct an instructive prompt.
|
39 |
+
prepend_prompt = "<LM>Проанализируй заданный ниже текст. Идентифицируй в нем грамматические и орфографические ошибки. Если есть хотя бы одна такая ошибка, то выведи 'True'. Если текст не содержит грамматических и орфографических ошибок, выведи 'False'.\n\nТекст: "
|
40 |
+
|
41 |
+
xx = ged_tokenizer([(prepend_prompt + input_text) for input_text in input_texts],
|
42 |
+
truncation=False,
|
43 |
+
padding="longest",
|
44 |
+
return_tensors='pt').to(ged_model.device)
|
45 |
+
out_ids = ged_model.generate(input_ids=xx.input_ids, eos_token_id=ged_tokenizer.eos_token_id, max_length=5)
|
46 |
+
|
47 |
+
# The model returns "True" or "False" text for each input row.
|
48 |
+
for input_text, has_defects in zip(input_texts, out_ids.cpu().tolist()):
|
49 |
+
has_defects = has_defects[1:has_defects.index(ged_tokenizer.eos_token_id)]
|
50 |
+
has_defects = ged_tokenizer.decode(has_defects)
|
51 |
+
print(f"{input_text} ==> {has_defects}")
|
52 |
+
```
|
53 |
+
|
54 |
+
The output should be:
|
55 |
+
|
56 |
+
```
|
57 |
+
Расскажу как настроить плагин и сделать быструю домашнюю страничку, расскажу как найти продвинутые гайды для создания сверх-эстетичной домашней страничке. ==> True
|
58 |
+
Расскажу, как настроить плагин и сделать быструю домашнюю страничку, расскажу, как найти продвинутые гайды для создания сверхэстетичной домашней странички. ==> False
|
59 |
+
```
|
60 |
+
|
61 |
+
## Metrics
|
62 |
+
|
63 |
+
The model was evaluated on the RUPOR dataset (which is not yet public) and on several open-source datasets:
|
64 |
+
|
65 |
+
| Domain | Population | F<sub>1</sub> | F<sub>0.5</sub> | Precision | Recall |
|
66 |
+
| ------------------------ | ---------- | ------------- | --------------- | --------- | ------ |
|
67 |
+
| RUPOR poetry | 4508 | 0.838 | 0.88 | 0.911 | 0.776 |
|
68 |
+
| RUPOR prose | 3998 | 0.882 | 0.911 | 0.932 | 0.837 |
|
69 |
+
| RuCoLa | 3998 | 0.268 | 0.464 | 0.905 | 0.158 |
|
70 |
+
| RuBLiMP | 3992 | 0.948 | 0.943 | 0.94 | 0.956 |
|
71 |
+
| rlc-toloka (ru) | 3992 | 0.801 | 0.846 | 0.878 | 0.737 |
|
72 |
+
| SAGE MultidomainGold | 2133 | 0.858 | 0.911 | 0.951 | 0.782 |
|
73 |
+
| SAGE RUSpellRU | 1233 | 0.941 | 0.963 | 0.979 | 0.906 |
|
74 |
+
| SAGE MedSpellchecker | 325 | 0.969 | 0.983 | 0.993 | 0.947 |
|
75 |
+
| SAGE GitHubTypoCorpusRu | 307 | 0.782 | 0.84 | 0.883 | 0.702 |
|
76 |
+
|
77 |
+
|