1 Hugging Face's logo
2 ---
3 language: rw
4 datasets:
5
6 ---
7 # bert-base-multilingual-cased-finetuned-kinyarwanda
8 ## Model description
9 **bert-base-multilingual-cased-finetuned-kinyarwanda** is a **Kinyarwanda BERT** model obtained by fine-tuning **bert-base-multilingual-cased** model on Kinyarwanda language texts. It provides **better performance** than the multilingual BERT on named entity recognition datasets.
10
11 Specifically, this model is a *bert-base-multilingual-cased* model that was fine-tuned on Kinyarwanda corpus.
12 ## Intended uses & limitations
13 #### How to use
14 You can use this model with Transformers *pipeline* for masked token prediction.
15 ```python
16 >>> from transformers import pipeline
17 >>> unmasker = pipeline('fill-mask', model='Davlan/bert-base-multilingual-cased-finetuned-kinyarwanda')
18 >>> unmasker("Twabonye ko igihe mu [MASK] hazaba hari ikirango abantu bakunze")
19
20 ```
21 #### Limitations and bias
22 This model is limited by its training dataset of entity-annotated news articles from a specific span of time. This may not generalize well for all use cases in different domains.
23 ## Training data
24 This model was fine-tuned on JW300 + [KIRNEWS](https://github.com/Andrews2017/KINNEWS-and-KIRNEWS-Corpus) + [BBC Gahuza](https://www.bbc.com/gahuza)
25
26 ## Training procedure
27 This model was trained on a single NVIDIA V100 GPU
28
29 ## Eval results on Test set (F-score, average over 5 runs)
30 Dataset| mBERT F1 | rw_bert F1
31 -|-|-
32 [MasakhaNER](https://github.com/masakhane-io/masakhane-ner) | 72.20 | 77.57
33
34 ### BibTeX entry and citation info
35 By David Adelani
36 ```
37
38 ```
39
40
41