This repository contains code for a sarcasm detector, which aims to classify headlines as sarcastic or non-sarcastic. The dataset used for training and evaluation consists of headlines along with their corresponding labels (sarcastic or not).

The dataset contains headlines with labels indicating whether each headline is sarcastic or not. The headlines will be tokenized and preprocessed to be fed into the model.

• The dataset is loaded from a JSON file using the pandas library.
• The headlines and labels are extracted from the dataset and split into training and testing sets using the specified split_size.
• The headlines are tokenized and converted to sequences using Tokenizer from tensorflow.keras.preprocessing.text.
• The sequences are then padded to a fixed length to ensure uniformity in input size.

• The sarcasm detector model is built using tensorflow.keras.models.Sequential.
• The model consists of an Embedding layer with a vocabulary size of 10,000 and an embedding dimension of 16. This layer converts the sequences of words into dense vectors.
• The GlobalAveragePooling1D layer averages the embeddings to obtain a fixed-length output.
• The model further contains two Dense layers with ReLU activation functions for feature extraction and a final Dense layer with a sigmoid activation function for binary classification (sarcastic or non-sarcastic).

• The model is trained using binary cross-entropy loss and the Adam optimizer for 30 epochs.
• The training and validation accuracy are monitored during training.

To use this sarcasm detector, you can follow these steps:

1. Install the required libraries and dependencies by running:

   pip install pandas numpy tensorflow

Feel free to experiment with different hyperparameters, model architectures, or other NLP techniques to improve the sarcasm detection accuracy.

For any questions or suggestions, please contact Francesco Alotto. Happy sarcasm detection with AI! 😄🤖