Add description for abbreviations
Browse files
README.md
CHANGED
@@ -11,8 +11,11 @@ widget:
|
|
11 |
## Model description
|
12 |
|
13 |
**CAMeLBERT** is a collection of BERT models pre-trained on Arabic texts with different sizes and variants.
|
|
|
|
|
14 |
The details are described in the paper *"[The Interplay of Variant, Size, and Task Type in Arabic Pre-trained Language Models](https://arxiv.org/abs/2103.06678)."*
|
15 |
-
|
|
|
16 |
|
17 |
||Model|Variant|Size|#Word|
|
18 |
|-|-|:-:|-:|-:|
|
@@ -25,8 +28,6 @@ We release eight models with different sizes and variants as follows:
|
|
25 |
||`bert-base-camelbert-msa-eighth`|MSA|14GB|1.6B|
|
26 |
||`bert-base-camelbert-msa-sixteenth`|MSA|6GB|746M|
|
27 |
|
28 |
-
This model card describes **CAMeLBERT-Mix** (`bert-base-camelbert-mix`), a model pre-trained on a mixture of these variants: CA, DA, and MSA.
|
29 |
-
|
30 |
## Intended uses
|
31 |
You can use the released model for either masked language modeling or next sentence prediction.
|
32 |
However, it is mostly intended to be fine-tuned on an NLP task, such as NER, POS tagging, sentiment analysis, dialect identification, and poetry classification.
|
@@ -83,15 +84,15 @@ output = model(encoded_input)
|
|
83 |
```
|
84 |
|
85 |
## Training data
|
86 |
-
- MSA
|
87 |
- [The Arabic Gigaword Fifth Edition](https://catalog.ldc.upenn.edu/LDC2011T11)
|
88 |
- [Abu El-Khair Corpus](http://www.abuelkhair.net/index.php/en/arabic/abu-el-khair-corpus)
|
89 |
- [OSIAN corpus](https://vlo.clarin.eu/search;jsessionid=31066390B2C9E8C6304845BA79869AC1?1&q=osian)
|
90 |
- [Arabic Wikipedia](https://archive.org/details/arwiki-20190201)
|
91 |
- The unshuffled version of the Arabic [OSCAR corpus](https://oscar-corpus.com/)
|
92 |
-
- DA
|
93 |
- A collection of dialectal Arabic data described in [our paper](https://arxiv.org/abs/2103.06678).
|
94 |
-
- CA
|
95 |
- [OpenITI (Version 2020.1.2)](https://zenodo.org/record/3891466#.YEX4-F0zbzc)
|
96 |
|
97 |
## Training procedure
|
|
|
11 |
## Model description
|
12 |
|
13 |
**CAMeLBERT** is a collection of BERT models pre-trained on Arabic texts with different sizes and variants.
|
14 |
+
We release pre-trained language models for Modern Standard Arabic (MSA), dialectal Arabic (DA), and classical Arabic (CA), in addition to a model pre-trained on a mix of the three.
|
15 |
+
We also provide additional models that are pre-trained on a scaled-down set of the MSA variant (half, quarter, eighth, and sixteenth).
|
16 |
The details are described in the paper *"[The Interplay of Variant, Size, and Task Type in Arabic Pre-trained Language Models](https://arxiv.org/abs/2103.06678)."*
|
17 |
+
|
18 |
+
This model card describes **CAMeLBERT-Mix** (`bert-base-camelbert-mix`), a model pre-trained on a mixture of these variants: MSA, DA, and CA.
|
19 |
|
20 |
||Model|Variant|Size|#Word|
|
21 |
|-|-|:-:|-:|-:|
|
|
|
28 |
||`bert-base-camelbert-msa-eighth`|MSA|14GB|1.6B|
|
29 |
||`bert-base-camelbert-msa-sixteenth`|MSA|6GB|746M|
|
30 |
|
|
|
|
|
31 |
## Intended uses
|
32 |
You can use the released model for either masked language modeling or next sentence prediction.
|
33 |
However, it is mostly intended to be fine-tuned on an NLP task, such as NER, POS tagging, sentiment analysis, dialect identification, and poetry classification.
|
|
|
84 |
```
|
85 |
|
86 |
## Training data
|
87 |
+
- MSA (Modern Standard Arabic)
|
88 |
- [The Arabic Gigaword Fifth Edition](https://catalog.ldc.upenn.edu/LDC2011T11)
|
89 |
- [Abu El-Khair Corpus](http://www.abuelkhair.net/index.php/en/arabic/abu-el-khair-corpus)
|
90 |
- [OSIAN corpus](https://vlo.clarin.eu/search;jsessionid=31066390B2C9E8C6304845BA79869AC1?1&q=osian)
|
91 |
- [Arabic Wikipedia](https://archive.org/details/arwiki-20190201)
|
92 |
- The unshuffled version of the Arabic [OSCAR corpus](https://oscar-corpus.com/)
|
93 |
+
- DA (dialectal Arabic)
|
94 |
- A collection of dialectal Arabic data described in [our paper](https://arxiv.org/abs/2103.06678).
|
95 |
+
- CA (classical Arabic)
|
96 |
- [OpenITI (Version 2020.1.2)](https://zenodo.org/record/3891466#.YEX4-F0zbzc)
|
97 |
|
98 |
## Training procedure
|