Downloads · 30 days
10
2% of all-time downloads
fahrendrakhoirul/indobert-cnn-finetuned-ecommerce-reviews
indobert-cnn-finetuned-ecommerce-reviews is a machine learning model from fahrendrakhoirul. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for transformers.
Short Summary: This model combines the power of IndoBERT for understanding language with a Convolutional Neural Network (CNN) for capturing local patterns in text data. It's specifically designed for multi-label class…
Downloads · 30 days
10
2% of all-time downloads
All-time downloads
519
Public
Parameters
126M
503 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors503 MB · 100%
From the Hugging Face model README
Short Summary:
This model combines the power of IndoBERT for understanding language with a Convolutional Neural Network (CNN) for capturing local patterns in text data. It's specifically designed for multi-label classification of customer reviews in e-commerce, focusing on: Product, Customer Sevice, and Shipping/Delivery
Detailed Description:
Explain that the model is based on IndoBERT-base-p1, a pre-trained IndoBERT model specifically designed for Indonesian text.
Highlight that it's fine-tuned on a dataset of e-commerce reviews, allowing it to understand the nuances of customer sentiment in this domain.
Following the IndoBERT layer, a 1D convolutional layer (Conv1d) is applied to extract local features from the hidden state vectors. An Adaptive Average Pooling layer (AdaptiveAvgPool1d) is used to reduce the dimensionality of the convolutional output, making it suitable for feeding into a final linear layer.
The final linear layer produces logits for each of the three classes (Produk, Layanan Pelanggan, Pengiriman). A sigmoid activation function is applied to convert the logits into probabilities, allowing for multi-label classification.
Clearly define the three output classes and their corresponding labels:
How to import in PyTorch:
import torch.nn as nn
from huggingface_hub import PyTorchModelHubMixin
from transformers import BertModel, AutoTokenizer
class IndoBertCNNEcommerceReview(nn.Module, PyTorchModelHubMixin):
def __init__(self, bert):
super().__init__()
self.bert = bert
self.conv1 = nn.Conv1d(in_channels=bert.config.hidden_size, out_channels=512, kernel_size=3, padding=1)
self.pool = nn.AdaptiveAvgPool1d(1)
self.linear = nn.Linear(512, 3)
self.sigmoid = nn.Sigmoid()
def forward(self, input_ids, attention_mask):
outputs = self.bert(input_ids=input_ids, attention_mask=attention_mask)
last_hidden_state = outputs.last_hidden_state
# Permute to [batch_size, hidden_size, seq_len]
last_hidden_state = last_hidden_state.permute(0, 2, 1)
conv1_output = self.conv1(last_hidden_state)
pooled_output = self.pool(conv1_output).squeeze(-1)
logits = self.linear(pooled_output)
probabilities = self.sigmoid(logits)
return probabilities
bert = BertModel.from_pretrained("indobenchmark/indobert-base-p1",
num_labels=3,
problem_type="multi_label_classification")
tokenizer = AutoTokenizer.from_pretrained("fahrendrakhoirul/indobert-cnn-finetuned-ecommerce-reviews")
model = IndoBertCNNEcommerceReview.from_pretrained("fahrendrakhoirul/indobert-cnn-finetuned-ecommerce-reviews", bert=bert)
This model has been pushed to the Hub using the PytorchModelHubMixin integration: