Basic Idea

• Train a statistical model on a corpus of documents with human-assigned titles.

• Compare 3 title generation methods:

  • Extractive Titles
  • Naïve Bayes
  • KNN

• Apply to machine translated documents

(2 * precision * recall)F1 = _________________ (precision + recall)

• Only measured word selection, not order

Previous slide

Next slide

Back to first slide

View graphic version