|
|
--- |
|
|
license: apache-2.0 |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- accuracy |
|
|
pipeline_tag: text-classification |
|
|
tags: |
|
|
- advertising |
|
|
--- |
|
|
|
|
|
# Tiny Bert Domain Advertising Classifier |
|
|
|
|
|
https://huggingface.co/ansi-code/bert-domain-advertising-classifier/blob/main/bert_domain_advertising_classifier.ipynb |
|
|
|
|
|
## Overview |
|
|
|
|
|
AdTargetingBERTClassifier is a small-scale BERT-based classifier designed for the task of ad targeting classification. The model is trained to predict multi-class labels associated with domains, as provided in the DAC693K dataset. |
|
|
|
|
|
## Model Architecture |
|
|
|
|
|
The classifier is built on the BERT (Bidirectional Encoder Representations from Transformers) architecture. It takes domain text as input and outputs logits for each class, enabling multi-class classification for ad targeting. |
|
|
|
|
|
## Model Training |
|
|
|
|
|
The model is trained on the "AdTargetingDataset" using a supervised learning approach. The training involves optimizing for the categorical cross-entropy loss, and the model is fine-tuned on the specific ad targeting classes associated with each domain. |
|
|
|
|
|
## Usage |
|
|
|
|
|
### Loading the Model |
|
|
|
|
|
To use the trained classifier in your Python environment, you can load it using the following code: |
|
|
|
|
|
```python |
|
|
from transformers import BertTokenizer, BertForSequenceClassification |
|
|
import torch |
|
|
|
|
|
# Load the pre-trained model and tokenizer |
|
|
model = BertForSequenceClassification.from_pretrained("ansi-code/bert-domain-advertising-classifier") |
|
|
tokenizer = BertTokenizer.from_pretrained("bert-base-uncased") |
|
|
|
|
|
# Example inference |
|
|
text = "google.com" |
|
|
inputs = tokenizer(text, return_tensors="pt") |
|
|
outputs = model(**inputs) |
|
|
logits = outputs.logits |
|
|
``` |
|
|
|
|
|
## Prediction |
|
|
To make predictions with the loaded model, you can use the obtained logits. Convert the logits to probabilities and determine the predicted class based on the highest probability. |
|
|
|
|
|
```python |
|
|
Copy code |
|
|
probabilities = torch.nn.functional.sigmoid(logits, dim=-1) |
|
|
predicted_class = torch.argmax(probabilities).item() |
|
|
``` |
|
|
|
|
|
## Model Evaluation |
|
|
|
|
|
The model's performance can be assessed using standard evaluation metrics such as accuracy, precision, recall, and F1-score on a separate validation set or through cross-validation. |
|
|
|
|
|
## License |
|
|
|
|
|
This model is released under the Apache 2.0 License. |
|
|
|
|
|
## Citation |
|
|
|
|
|
If you use this model in your work, please cite it using the following BibTeX entry: |
|
|
|
|
|
```bibtex |
|
|
@model{silvi_2023_bert-domain-advertising-classifier, |
|
|
title = {bert-domain-advertising-classifier}, |
|
|
author = {Andrea Silvi}, |
|
|
year = {2023}, |
|
|
} |
|
|
``` |
|
|
|
|
|
## Acknowledgements |
|
|
|
|
|
We would like to thank the developers of the Hugging Face Transformers library for providing the BERT model implementation. |
|
|
|
