xlm-roberta-large-english-party-cap-v3

Model description

An xlm-roberta-large model finetuned on norwegian training data containing texts of the party domain labelled with major topic codes from the Comparative Agendas Project.

How to use the model

Loading and tokenizing input data

import pandas as pd
import numpy as np
from datasets import Dataset
from transformers import (AutoModelForSequenceClassification, AutoTokenizer, 
                          Trainer, TrainingArguments)

CAP_NUM_DICT = {0: '1', 1: '2', 2: '3', 3: '4', 4: '5', 5: '6', 
6: '7', 7: '8', 8: '9', 9: '10', 10: '12', 11: '13', 12: '14', 
13: '15', 14: '16', 15: '17', 16: '18', 17: '19', 18: '20', 19: 
'21', 20: '23', 21: '999'}

tokenizer = AutoTokenizer.from_pretrained('xlm-roberta-large')
num_labels = len(CAP_NUM_DICT)

def tokenize_dataset(data : pd.DataFrame):
    tokenized = tokenizer(data["text"],
                          max_length=256,
                          truncation=True,
                          padding="max_length")
    return tokenized

hg_data = Dataset.from_pandas(data)
dataset = hg_data.map(tokenize_dataset, batched=True, remove_columns=hg_data.column_names)

Inference using the Trainer class

model = AutoModelForSequenceClassification.from_pretrained('poltextlab/xlm-roberta-large-norwegian-party-cap-v3',
                                                           num_labels=num_labels,
                                                           problem_type="multi_label_classification",
                                                           ignore_mismatched_sizes=True
                                                           )

training_args = TrainingArguments(
    output_dir='.',
    per_device_train_batch_size=32,
    per_device_eval_batch_size=32
)

trainer = Trainer(
    model=model,
    args=training_args
)

probs = trainer.predict(test_dataset=dataset).predictions
predicted = pd.DataFrame(np.argmax(probs, axis=1)).replace({0: CAP_NUM_DICT}).rename(
    columns={0: 'predicted'}).reset_index(drop=True)

Fine-tuning procedure

xlm-roberta-large-norwegian-party-cap-v3 was fine-tuned using the Hugging Face Trainer class with the following hyperparameters:

training_args = TrainingArguments(
    output_dir=f"../model/{model_dir}/tmp/",
    logging_dir=f"../logs/{model_dir}/",
    logging_strategy='epoch',
    num_train_epochs=10,
    per_device_train_batch_size=32,
    per_device_eval_batch_size=32,
    learning_rate=5e-06,
    seed=42,
    save_strategy='epoch',
    evaluation_strategy='epoch',
    save_total_limit=1,
    load_best_model_at_end=True
)

We also incorporated an EarlyStoppingCallback in the process with a patience of 2 epochs.

Model performance

The model was evaluated on a test set of 21915 examples (20% of the available data).
Model accuracy is 0.83.

label	precision	recall	f1-score	support
0	0.78	0.78	0.78	489
1	0.83	0.84	0.83	624
2	0.87	0.87	0.87	868
3	0.83	0.86	0.85	473
4	0.80	0.79	0.80	527
5	0.92	0.89	0.90	831
6	0.88	0.80	0.84	932
7	0.83	0.87	0.85	320
8	0.89	0.83	0.86	327
9	0.86	0.88	0.87	466
10	0.90	0.87	0.89	621
11	0.76	0.80	0.78	766
12	0.77	0.81	0.79	289
13	0.75	0.75	0.75	579
14	0.85	0.90	0.88	319
15	0.81	0.80	0.80	328
16	0.73	0.70	0.72	146
17	0.83	0.82	0.83	708
18	0.75	0.80	0.77	582
19	0.80	0.74	0.77	188
20	0.81	0.85	0.83	233
21	0.90	0.95	0.92	342
accuracy	0.83	0.83	0.83	0.83
macro avg	0.83	0.83	0.83	10958
weighted avg	0.83	0.83	0.83	10958

Inference platform

This model is used by the CAP Babel Machine, 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.

Debugging and issues

This architecture uses the sentencepiece tokenizer. In order to run 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.

poltextlab
/

xlm-roberta-large-norwegian-party-cap-v3

You need to agree to share your contact information to access this model