CNN for Computer Vision

CNNs are most well-known for solving Computer Vision problems because they can reliably detect patterns in images.

Embedding the Article

We embed the text in a way that maps related words to "nearby" vectors. Then, articles that stay on topic tend to span a particular region.

Applying the CNN to NLP

The CNN can recognize patterns in the embeddings like how it would recognize edges or shapes in images. It uses the patterns to detect if the article is real!

Using the CNN, we reached the following results:

Test Accuracy: 91.38%

Training Accuracy: 93%