Using Local Alignments for Relation Recognition

TL;DR

This paper introduces local alignment kernels based on Smith-Waterman scores to improve relation recognition in text, effectively combining structural and semantic information, and demonstrates promising results on biomedical and general datasets.

Contribution

The paper proposes a novel local alignment kernel for relation recognition that integrates structural similarity with semantic knowledge, outperforming baseline methods.

Findings

LA kernel outperforms baselines on biomedical corpora

Performance is comparable to state-of-the-art on general relation datasets

Incorporating distributional similarity enhances relation recognition

Abstract

This paper discusses the problem of marrying structural similarity with semantic relatedness for Information Extraction from text. Aiming at accurate recognition of relations, we introduce local alignment kernels and explore various possibilities of using them for this task. We give a definition of a local alignment (LA) kernel based on the Smith-Waterman score as a sequence similarity measure and proceed with a range of possibilities for computing similarity between elements of sequences. We show how distributional similarity measures obtained from unlabeled data can be incorporated into the learning task as semantic knowledge. Our experiments suggest that the LA kernel yields promising results on various biomedical corpora outperforming two baselines by a large margin. Additional series of experiments have been conducted on the data sets of seven general relation types, where the performance of the LA kernel is comparable to the current state-of-the-art results.