Liuliu writeup of Cohen 2000 AI

This is a review of Cohen_2000_whirl_a_word_based_information_representation_language by user:Liuliu.

This paper introduces a data representation language WHIRL which combines soft logic with textual similarities. It can do deduction with approximate matching, returning the highest scored binding of variables of user's query to user. The efficiency is assured by using a sort-of beam search method to only find top k highest substitutions.

I was wondering whether Markov logic learns something from this representation method: both are logic with some numeric scores.