Edit model card

dipawidia/xlnet-base-cased-product-review-sentiment-analysis

This model is a fine-tuned version of xlnet-base-cased on any type of product reviews dataset gathered from several e-commerce such as shopee, tokopedia, blibli, lazada, and zalora. The dataset can be found here It achieves the following results on the evaluation set:

  • Train Loss: 0.1085
  • Train Accuracy: 0.9617
  • Validation Loss: 0.1910
  • Validation Accuracy: 0.9414
  • Epoch: 4

Intended uses & limitations

This fine-tuned XLNet model is used for sentiment analysis with 2 labels text classification: 0 -> Negative; 1 -> Positive.

Example Pipeline

from transformers import pipeline
pipe = pipeline("text-classification", model="dipawidia/xlnet-base-cased-product-review-sentiment-analysis")
pipe("This shoes is awesome")
[{'label': 'Positive', 'score': 0.9995703101158142}]

Full classification example

from transformers import XLNetTokenizer, TFXLNetForSequenceClassification
import tensorflow as tf
import numpy as np

tokenizer = XLNetTokenizer.from_pretrained("dipawidia/xlnet-base-cased-product-review-sentiment-analysis")
model = TFXLNetForSequenceClassification.from_pretrained("dipawidia/xlnet-base-cased-product-review-sentiment-analysis")

def get_sentimen(text):
  tokenize_text = tokenizer(text, return_tensors = 'tf')
  preds = model.predict(dict(tokenize_text))['logits']
  class_preds = np.argmax(tf.keras.layers.Softmax()(preds))
  if class_preds == 1:
    label = 'Positive'
  else:
    label = 'Negative'
  return(label)

get_sentimen('i hate this product')

Output:

Negative

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • optimizer: {'name': 'AdamW', 'weight_decay': 0.004, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': 3e-05, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
  • training_precision: float32

Training results

Train Loss Train Accuracy Validation Loss Validation Accuracy Epoch
0.3417 0.8491 0.1568 0.9449 0
0.1943 0.9235 0.1504 0.9466 1
0.1569 0.9404 0.1612 0.9466 2
0.1238 0.9572 0.1748 0.9475 3
0.1085 0.9617 0.1910 0.9414 4

Framework versions

  • Transformers 4.41.2
  • TensorFlow 2.15.0
  • Tokenizers 0.19.1
Downloads last month
14
Safetensors
Model size
117M params
Tensor type
F32
·

Finetuned from

Dataset used to train dipawidia/xlnet-base-cased-product-review-sentiment-analysis