Update README.md
Browse files
README.md
CHANGED
@@ -1,26 +1,34 @@
|
|
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 |
For more resources, including the source code, please refer to the GitHub repository [gohjiayi/suicidal-text-detection](https://github.com/gohjiayi/suicidal-text-detection/).
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
base_model:
|
6 |
+
- google-bert/bert-base-uncased
|
7 |
+
pipeline_tag: text-classification
|
8 |
+
---
|
9 |
+
# Suicidal-BERT
|
10 |
+
This text classification model predicts whether a sequence of words are suicidal (1) or non-suicidal (0).
|
11 |
+
|
12 |
+
## Data
|
13 |
+
The model was trained on the [Suicide and Depression Dataset](https://www.kaggle.com/nikhileswarkomati/suicide-watch) obtained from Kaggle. The dataset was scraped from Reddit and consists of 232,074 rows equally distributed between 2 classes - suicide and non-suicide.
|
14 |
+
|
15 |
+
## Parameters
|
16 |
+
The model fine-tuning was conducted on 1 epoch, with batch size of 6, and learning rate of 0.00001. Due to limited computing resources and time, we were unable to scale up the number of epochs and batch size.
|
17 |
+
|
18 |
+
## Performance
|
19 |
+
The model has achieved the following results after fine-tuning on the aforementioned dataset:
|
20 |
+
- Accuracy: 0.9757
|
21 |
+
- Recall: 0.9669
|
22 |
+
- Precision: 0.9701
|
23 |
+
- F1 Score: 0.9685
|
24 |
+
|
25 |
+
## How to Use
|
26 |
+
Load the model via the transformers library:
|
27 |
+
```
|
28 |
+
from transformers import AutoTokenizer, AutoModel
|
29 |
+
tokenizer = AutoTokenizer.from_pretrained("gooohjy/suicidal-bert")
|
30 |
+
model = AutoModel.from_pretrained("gooohjy/suicidal-bert")
|
31 |
+
```
|
32 |
+
|
33 |
+
## Resources
|
34 |
For more resources, including the source code, please refer to the GitHub repository [gohjiayi/suicidal-text-detection](https://github.com/gohjiayi/suicidal-text-detection/).
|