People sometimes want to know a specific reason for a page classification. For example, was the content category “Infectious Diseases” listed because the word “Cold” appeared on the page?
The answer to the question lies in a thousand different numbers in a neural network. The numbers are tweaked as the model dissects innumerable pieces of training data, picking up on subtleties that may not be immediately obvious to humans. For example, a neural network can recognize:
Positive or negative words
Long or short sentences or paragraphs
FIrst, second, or third-person voice
Density of words of similar meaning
Certain words coinciding with certain images
The power of a neural network is manifest in its ability to recognize hidden patterns and correlations in raw data, cluster and classify the data, and continuously learn and improve over time.