This model is a fine-tuned version of XLM-RoBERTa (XLM-R) specifically designed to handle the complexities of Code-Mixed Minangkabau sentiment analysis. Social media data in Indonesia, particularly from regional areas like West Sumatra, often features a heavy mix of Minangkabau, Indonesian, and English.

Traditional models often struggle with this "low-resource gap" and "non-standard orthography." This model bridge those gaps while prioritizing Explainability (XAI) to demystify the "black-box" nature of Deep Learning.

Key Features:

Multilingual Support: Optimized for Indonesian, Minangkabau, and English code-mixing.
Robustness: Handles informal spelling, unique affixations, and phonetic typing common in YouTube comments.
XAI-Ready: Designed to be interpreted using feature attribution methods like SHAP or LIME to provide local and global explanations.

Intended Uses & Limitations

Primary Use: Sentiment classification (Positive, Neutral, Negative) for regional Indonesian languages.
Limitations: Performance might vary on purely formal Minang literature as the training data is derived from social media (YouTube) contexts.

How to Use

You can use this model directly with the transformers library:

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "bimobirra/explainable-xlmr-code-mixed-low-resource-lang"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

text = "Rancak bana videonya, tapi agak sedikit nge-lag loadingnya."
inputs = tokenizer(text, return_tensors="pt")

with torch.no_grad():
    logits = model(**inputs).logits

predicted_class_id = logits.argmax().item()
print(f"Predicted class ID: {predicted_class_id}")