Comparing traditional text classification and deep learning

The traditional text classification performed a number of preprocessing steps, including word stemming, stop-word processing, and feature generation (tf-idf, tf or binary). The deep learning text classification did not need this preprocessing. You may have heard various reasons for this previously:

  • Deep learning can learn features automatically, so feature creation is not needed
  • Deep learning algorithms for NLP tasks requires far less preprocessing than traditional text classification

There is some truth to this, but this does not answer why we need complex feature generation in traditional text classification. A big reason that preprocessing is needed in traditional text classification ...

Get R Deep Learning Essentials now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.