arnolfokam commited on
Commit
f1b4508
1 Parent(s): 7b1e892

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +45 -1
README.md CHANGED
@@ -14,4 +14,48 @@ widget:
14
  - text: "Ambasaderi Bellomo yavuze ko bishimira ubufatanye burambye hagati ya EU n’u Rwanda, bushingiye nanone ku bufatanye hagati y’imigabane ya Afurika n’u Burayi."
15
  ---
16
 
17
- Model description
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  - text: "Ambasaderi Bellomo yavuze ko bishimira ubufatanye burambye hagati ya EU n’u Rwanda, bushingiye nanone ku bufatanye hagati y’imigabane ya Afurika n’u Burayi."
15
  ---
16
 
17
+ # Model description
18
+ **bert-base-cased-kin** is a model based on the fine-tuned BERT base cased model. It has been trained to recognize four types of entities:
19
+
20
+ - dates & time (DATE)
21
+ - Location (LOC)
22
+ - Organizations (ORG)
23
+ - Person (PER)
24
+
25
+ # Intended Use
26
+ #### HowTo
27
+ - Intended to be used for research purposes concerning Named Entity Recognition for African Languages.
28
+ - Not intended for practical purposes.
29
+
30
+ # Training Data
31
+ This model was fine-tuned on the Kinyarwanda corpus **(kin)** of the [MasakhaNER](https://github.com/masakhane-io/masakhane-ner) dataset. However, we thresholded the number of entity groups per sentence in this dataset to 10 entity groups.
32
+
33
+ # Training procedure
34
+ This model was trained on a single NVIDIA P5000 from [Paperspace](https://www.paperspace.com)
35
+ #### Hyperparameters
36
+ - **Learning Rate:** 5e-5
37
+ - **Batch Size:** 32
38
+ - **Maximum Sequence Length:** 164
39
+ - **Epochs:** 30
40
+
41
+
42
+ # Evaluation Data
43
+ We evaluated this model on the test split of the Kinyarwandan corpus **(kin)** present in the [MasakhaNER](https://github.com/masakhane-io/masakhane-ner) with no thresholding.
44
+
45
+ # Metrics
46
+ - Precision
47
+ - Recall
48
+ - F1-score
49
+
50
+ # Limitations
51
+ - The size of the pre-trained language model prevents its usage in anything other than research.
52
+ - Lack of analysis concerning the bias and fairness in these models may make them dangerous if deployed into production system.
53
+ - The train data is a less populated version of the original dataset in terms of entity groups per sentence. Therefore, this can negatively impact the performance.
54
+
55
+ # Caveats and Recommendations
56
+ - The topics in the dataset corpus are centered around **News**. Future training could be done with a more diverse corpus.
57
+
58
+ # Results
59
+ Model Name| Precision | Recall | F1-score
60
+ -|-|-|-
61
+ **bert-base-cased-kin**| 75.00 |80.09|77.47