spam-mail-detection-w-tensorflow-distilbert

Spam Mail Detection w/Tensorflow (DistilBERT)

(kaggle link -> https://www.kaggle.com/code/banddaniel/spam-mail-detection-w-tensorflow-distilbert)

I tried to predict a spam mail with finetuning a DistilBert based Tensorflow model.

• I applied several preprocessing operations (cleaning,dropping stop words),
• Used tf.data pipeline for efficient training,
• I only used only 20 max length for sequence length (bert models support up to 512 input lengths),
• Only 18000 samples be used for training (12000 samples for validating and 20000 samples for testing),

My Another Projects

• Complaint Analysis w/Ensemble Model (CatBoost, LR)
• Gemma 2B Text Summarization w/Zero-Shot Prompting
• Rating Prediction w/SentenceTransformer, CatBoost

References

1. https://towardsdatascience.com/hugging-face-transformers-fine-tuning-distilbert-for-binary-classification-tasks-490f1d192379
2. https://www.kaggle.com/code/preatcher/emotion-detection-by-using-bert