Unsupervised Open Relation Extraction

TL;DR

This paper presents an unsupervised method for extracting relations between named entities from text, introducing a novel word embedding re-weighting technique and feature reduction to improve clustering performance.

Contribution

It proposes a new unsupervised relation extraction approach with a novel word embedding re-weighting and feature reduction, achieving significant performance gains.

Findings

Improved F1-score of 0.416 on NYT-FB dataset

Achieved 5.8% improvement over state-of-the-art

Effective handling of feature sparsity

Abstract

We explore methods to extract relations between named entities from free text in an unsupervised setting. In addition to standard feature extraction, we develop a novel method to re-weight word embeddings. We alleviate the problem of features sparsity using an individual feature reduction. Our approach exhibits a significant improvement by 5.8% over the state-of-the-art relation clustering scoring a F1-score of 0.416 on the NYT-FB dataset.