Reducing Semantic Drift with Bagging and Distributional Similarity

Iterative bootstrapping algorithms are typically compared using a single set of handpicked seeds. However, we demonstrate that performance varies greatly depending on these seeds, and favourable seeds for one algorithm can perform very poorly with others, making comparisons unreliable. We exploit this wide variation with bagging, sampling from automatically extracted seeds to reduce semantic drift. However, semantic drift still occurs in later iterations. We propose an integrated distributional similarity filter to identify and censor potential semantic drifts, ensuring over 10% higher precision when extracting large semantic lexicons.

• Type: Journal
• Year: 2009
• Where: ACL
• Authors: Tara McIntosh, James R. Curran
• Keywords: ACL 2009, Computational Linguistics, Iterative Bootstrapping Algorithms, Potential Semantic Drifts, Semantic Drifts