|
|
|
--- |
|
--- |
|
language: |
|
- nl |
|
tags: |
|
- text-classification |
|
- pytorch |
|
metrics: |
|
- accuracy |
|
- f1-score |
|
--- |
|
# xlm-roberta-large-dutch-execspeech-cap-v3 |
|
## Model description |
|
An `xlm-roberta-large` model fine-tuned on dutch training data containing executive speeches labeled with [major topic codes](https://www.comparativeagendas.net/pages/master-codebook) from the [Comparative Agendas Project](https://www.comparativeagendas.net/). |
|
|
|
## How to use the model |
|
This snippet prints the three most probable labels and their corresponding softmax scores: |
|
```python |
|
import torch |
|
from transformers import AutoModelForSequenceClassification, AutoTokenizer |
|
|
|
model = AutoModelForSequenceClassification.from_pretrained("poltextlab/xlm-roberta-large-dutch-execspeech-cap-v3") |
|
tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-large") |
|
|
|
sentence = "This is an example." |
|
|
|
inputs = tokenizer(sentence, |
|
return_tensors="pt", |
|
max_length=512, |
|
padding="do_not_pad", |
|
truncation=True |
|
) |
|
|
|
logits = model(**inputs).logits |
|
|
|
probs = torch.softmax(logits, dim=1).tolist()[0] |
|
probs = {model.config.id2label[index]: round(probability, 2) for index, probability in enumerate(probs)} |
|
top3_probs = dict(sorted(probs.items(), key=lambda item: item[1], reverse=True)[:3]) |
|
|
|
print(top3_probs) |
|
``` |
|
|
|
## Model performance |
|
The model was evaluated on a test set of 897 examples.<br> |
|
Model accuracy is **0.67**. |
|
| label | precision | recall | f1-score | support | |
|
|:-------------|------------:|---------:|-----------:|----------:| |
|
| 0 | 0.71 | 0.73 | 0.72 | 248 | |
|
| 1 | 0 | 0 | 0 | 8 | |
|
| 2 | 0.88 | 0.81 | 0.84 | 26 | |
|
| 3 | 0.67 | 0.29 | 0.4 | 7 | |
|
| 4 | 0.58 | 0.78 | 0.67 | 128 | |
|
| 5 | 0 | 0 | 0 | 2 | |
|
| 6 | 0.73 | 0.73 | 0.73 | 11 | |
|
| 7 | 0.67 | 0.57 | 0.62 | 14 | |
|
| 8 | 0.88 | 0.71 | 0.79 | 21 | |
|
| 9 | 0.78 | 0.78 | 0.78 | 18 | |
|
| 10 | 0.78 | 0.73 | 0.75 | 115 | |
|
| 11 | 0.51 | 0.56 | 0.53 | 59 | |
|
| 12 | 1 | 0.2 | 0.33 | 5 | |
|
| 13 | 0.52 | 0.34 | 0.41 | 38 | |
|
| 14 | 0.47 | 0.56 | 0.51 | 16 | |
|
| 15 | 0.8 | 0.5 | 0.62 | 8 | |
|
| 16 | 0.3 | 0.3 | 0.3 | 20 | |
|
| 17 | 0.71 | 0.66 | 0.68 | 73 | |
|
| 18 | 0.67 | 0.62 | 0.65 | 80 | |
|
| 19 | 0 | 0 | 0 | 0 | |
|
| macro avg | 0.58 | 0.49 | 0.52 | 897 | |
|
| weighted avg | 0.67 | 0.67 | 0.66 | 897 | |
|
|
|
### Fine-tuning procedure |
|
This model was fine-tuned with the following key hyperparameters: |
|
|
|
- **Number of Training Epochs**: 10 |
|
- **Batch Size**: 8 |
|
- **Learning Rate**: 5e-06 |
|
- **Early Stopping**: enabled with a patience of 2 epochs |
|
|
|
## Inference platform |
|
This model is used by the [CAP Babel Machine](https://babel.poltextlab.com), an open-source and free natural language processing tool, designed to simplify and speed up projects for comparative research. |
|
|
|
## Cooperation |
|
Model performance can be significantly improved by extending our training sets. We appreciate every submission of CAP-coded corpora (of any domain and language) at poltextlab{at}poltextlab{dot}com or by using the [CAP Babel Machine](https://babel.poltextlab.com). |
|
|
|
## Reference |
|
Sebők, M., Máté, Á., Ring, O., Kovács, V., & Lehoczki, R. (2024). Leveraging Open Large Language Models for Multilingual Policy Topic Classification: The Babel Machine Approach. Social Science Computer Review, 0(0). https://doi.org/10.1177/08944393241259434 |
|
|
|
## Debugging and issues |
|
This architecture uses the `sentencepiece` tokenizer. In order to use the model before `transformers==4.27` you need to install it manually. |
|
|
|
If you encounter a `RuntimeError` when loading the model using the `from_pretrained()` method, adding `ignore_mismatched_sizes=True` should solve the issue. |
|
|
