[ Case study ]

Echo Truth

Machine-learning fake news detection

PythonNLPMachine LearningKaggle

Overview

Echo Truth is a machine-learning fake news detector. It scrutinizes news content and verifies its authenticity. In a landscape where misinformation can sway public opinion in hours, the goal was a dependable model that helps readers navigate media with confidence and discernment.

Approach

01Assembled a comprehensive labeled news dataset from Kaggle, balancing real and fabricated sources.
02Cleaned and vectorized article text: tokenization, stop-word handling, and TF-IDF feature extraction.
03Trained and compared multiple classifiers, tuning for precision so genuine reporting is never casually flagged.
04Evaluated on held-out data, then ran error analysis on the hardest misclassified examples to guide iteration.

Results

Pipeline: how it actually works

Kaggle news dataset

Labeled real and fabricated articles, balanced for training.

Clean & vectorize

Tokenization, stop-word handling, and TF-IDF feature extraction.

Train classifiers

Multiple models trained and compared, tuned for precision.

Evaluate & analyze errors

Held-out evaluation plus review of the hardest misclassifications.

Error analysis showed opinion pieces sit on the hardest decision boundary, flagged for future work.

View the repository