Commit
•
bc1988f
1
Parent(s):
ef5ed2c
Update README.md
Browse files
README.md
CHANGED
@@ -4,4 +4,36 @@ language:
|
|
4 |
- en
|
5 |
- fr
|
6 |
- es
|
7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
- en
|
5 |
- fr
|
6 |
- es
|
7 |
+
- multilingual
|
8 |
+
---
|
9 |
+
|
10 |
+
# HumBert
|
11 |
+
|
12 |
+
HumBert is a [XLM-Roberta](https://huggingface.co/xlm-roberta-base) model trained on humanitarian texts - approximately 50 million textual examples (roughly 2 billion tokens) from public humanitarian reports, law cases and news articles.
|
13 |
+
Data were collected from three main sources: [Reliefweb](https://reliefweb.int/), [UNHCR Refworld](https://www.refworld.org/) and [Europe Media Monitor News Brief](https://emm.newsbrief.eu/).
|
14 |
+
Although XLM-Roberta was trained on 100 different languages, this fine-tuning was performed on three languages, English, French and Spanish, due to the impossibility of finding a good amount of such kind of humanitarian data in other languages.
|
15 |
+
|
16 |
+
|
17 |
+
## Intended uses & limitations
|
18 |
+
|
19 |
+
To the best of our knowledge, HumBert is the first language model adapted on humanitarian topics, which often use a very specific language, making adaptation to downstream tasks (such as dister responses text classification) more effective.
|
20 |
+
This model is primarily aimed at being fine-tuned on tasks such as sequence classification or token classification.
|
21 |
+
|
22 |
+
## Benchmarks
|
23 |
+
|
24 |
+
Soon...
|
25 |
+
|
26 |
+
## Usage
|
27 |
+
|
28 |
+
Here is how to use this model to get the features of a given text in PyTorch:
|
29 |
+
|
30 |
+
```python
|
31 |
+
from transformers import AutoTokenizer, AutoModelForMaskedLM
|
32 |
+
tokenizer = AutoTokenizer.from_pretrained('nlp-thedeep/humbert')
|
33 |
+
model = AutoModelForMaskedLM.from_pretrained("nlp-thedeep/humbert")
|
34 |
+
# prepare input
|
35 |
+
text = "Replace me by any text you'd like."
|
36 |
+
encoded_input = tokenizer(text, return_tensors='pt')
|
37 |
+
# forward pass
|
38 |
+
output = model(**encoded_input)
|
39 |
+
```
|