Directed Research Project Details
Refining NELL's knowledge base with Markov logic networks
| Author: | Shangpu Jiang |
|---|---|
| Date: | December 06, 2011 |
| Time: | 9:30 |
| Location: | 220 Deschutes |
| Committee: | Dejing Dou (Chair) Daniel Lowd Arthur Farley |
Abstract
Never Ending Language Learner (NELL) is an AI system that runs 24 hours per day, 7 days per week, forever, repeatly extracting knowledge from the web. It uses a bootstrap learning algorithm that works with small volume of labeled data, and large volume of unlabeled data. One of the biggest problems of NELL is that the accuracy of the knowledge it acquires drops down over time gradually. In this work, we propose a novel approach dedicated to solving the problem. Our approach uses Markov Logic Networks to remodel NELL's knowledge base, and conducts an inference algorithm on it to find the true knowledge out of all the candidate knowledge NELL generates. The preliminary experiment shows that compared with NELL's own heuristic by which selective knowledge is promoted into the knowledge base, our method achieves better recall and F1 score (harmonic mean of precision and recall), while maintaining reasonable precision. Our method creates a new perspective towards which traditional first order logic information extraction system can be improved.