Semantic Relation Classification via Convolutional Neural Networks with Simple Negative Sampling

TL;DR

This paper introduces a convolutional neural network approach focusing on shortest dependency paths and a negative sampling strategy to improve semantic relation classification, achieving superior results on a benchmark dataset.

Contribution

It presents a novel CNN-based method utilizing shortest dependency paths and a simple negative sampling technique for more accurate relation classification.

Findings

Outperforms state-of-the-art on SemEval-2010 Task 8

Effective use of shortest dependency paths in CNNs

Negative sampling improves relation assignment accuracy

Abstract

Syntactic features play an essential role in identifying relationship in a sentence. Previous neural network models often suffer from irrelevant information introduced when subjects and objects are in a long distance. In this paper, we propose to learn more robust relation representations from the shortest dependency path through a convolution neural network. We further propose a straightforward negative sampling strategy to improve the assignment of subjects and objects. Experimental results show that our method outperforms the state-of-the-art methods on the SemEval-2010 Task 8 dataset.