ZycckZ/XLM-Align-base_ABSA_finetuned

XLM-Align Base for Aspect-Based Sentiment Analysis (Vietnamese)

1. Model Description

This model is a fine-tuned version of microsoft/xlm-align-base for the task of Aspect-Based Sentiment Analysis (ABSA) on Vietnamese text. It is capable of identifying sentiments (Positive, Neutral, Negative) for specific aspects within a given sentence.

• Language: Vietnamese
• Model Type: Transformer-based Encoder
• Base Model: microsoft/xlm-align-base

2. Training Data

The model was trained on Vietnamese ABSA datasets including thinhntr/absa and LuongPhan/UIT-ViSFD.

3. Training Procedure

Hyperparameters:

• Learning Rate: 5e-5
• Batch Size: 32
• Precision: Mixed precision (fp16)
• Epochs: 3

4. Evaluation Results

The model was evaluated on the test set (9141 samples).

• Accuracy: 84.42%
• F1-Macro: 76.78%

Classification Report:

Class	Precision	Recall	F1-Score
0 (Negative)	0.8250	0.8734	0.8485
1 (Neutral)	0.7098	0.4517	0.5521
2 (Positive)	0.8736	0.9344	0.9030

5. How to Use

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("ZycckZ/XLM-Align-base_ABSA_finetuned")
model = AutoModelForSequenceClassification.from_pretrained("ZycckZ/XLM-Align-base_ABSA_finetuned")