File size: 1,803 Bytes
33d59f7
c98208a
8f53ade
 
33d59f7
74ab5eb
ba8448b
74ab5eb
 
 
 
 
 
9977692
74ab5eb
0da8032
 
 
 
 
 
 
74ab5eb
 
 
4cc3454
 
0c0a1dc
 
8daf32d
 
 
 
4cc3454
 
0c0a1dc
4cc3454
 
 
 
 
 
 
 
 
0c0a1dc
 
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
---
pipeline_tag: fill-mask
widget:
 - text: "đậu xanh rau <mask>"
---
# <a name="introduction"></a> ViSoBERT: A Pre-Trained Language Model for Vietnamese Social Media Text Processing (EMNLP 2023 - Main)
**Disclaimer**: The paper contains actual comments on social networks that might be construed as abusive, offensive, or obscene.

ViSoBERT is the state-of-the-art language model for Vietnamese social media tasks:

 - ViSoBERT is the first monolingual MLM (XLM-R architecture) from scratch specifically for Vietnamese social media text.
 - ViSoBERT outperforms previous monolingual, multilingual, and multilingual social media approaches, obtaining new state-of-the-art performances on four downstream Vietnamese social media tasks.

The general architecture and experimental results of ViSoBERT can be found in our [paper](https://arxiv.org/abs/2310.11166):

    @misc{nguyen2023visobert,
          title={ViSoBERT: A Pre-Trained Language Model for Vietnamese Social Media Text Processing}, 
          author={Quoc-Nam Nguyen and Thang Chau Phan and Duc-Vu Nguyen and Kiet Van Nguyen},
          year={2023},
          eprint={2310.11166},
          archivePrefix={arXiv},
          primaryClass={cs.CL}
    }
    

**Please CITE** our paper when ViSoBERT is used to help produce published results or is incorporated into other software.

**Installation** 

Install `transformers` and `SentencePiece` packages:
    
    pip install transformers
    pip install SentencePiece

**Example usage**
```python
from transformers import AutoModel,AutoTokenizer
import torch

model= AutoModel.from_pretrained('uitnlp/visobert')
tokenizer = AutoTokenizer.from_pretrained('uitnlp/visobert')

encoding = tokenizer('dau xanh rau ma',return_tensors='pt')

with torch.no_grad():
  output = model(**encoding)
```