Sgopal1 writeup of WHIRL

This is a review of the paper Cohen_2000_whirl_a_word_based_information_representation_language by user:sgopal1.

This paper proposes a an algorithm, syntax and structure for retrieving information from structured data. It looks extremely similar to prolog, but uses difference inference mechanisms to retrieve the r-best-lists ( r-materialization ).

Key points

• Supports conjunctive queries.
• Soft-matching between texts using TF-IDF or any other similarity metric.
• The database of facts along with scores, appropriately combines scores and presents it to the user

Applied WHIRL to three areas

• Ranking
• Classification
• Extraction from structured documents like HTML