Sarcasm Detection App

Overview

Sarcasm is a small language problem with a big user experience effect. This app explores how an LSTM-based text classifier can move from dataset training into a simple interface that lets people test predictions in real time.

Problem

Many ML projects stop at model output, but I wanted to connect training work to an experience that non-technical users could interact with directly. The challenge was to make the workflow feel practical, not purely academic.

What I built

• dataset ingestion and preprocessing for sarcasm-labelled text
• model training for binary classification
• a Streamlit interface for immediate text-based inference
• a repo structure that supports training, saved models, and local experimentation

Product thinking

The interesting part of this project is the bridge between model development and usability. A prediction model is much more compelling when users can actually test it, understand it, and see the value immediately.

Outcome

This project showcases both NLP experimentation and the ability to package ML work into a shareable interface. It also leaves room for future improvements such as richer explainability, confidence scores, or alternative language models.