Luminoso at SemEval-2018 Task 10: Distinguishing Attributes Using Text Corpora and Relational Knowledge

TL;DR

This paper describes Luminoso's system for SemEval-2018 that uses ConceptNet and a linear classifier to identify discriminative attributes in text, achieving near-top performance.

Contribution

The paper introduces a method combining ConceptNet-based features with a linear classifier for attribute discrimination, demonstrating competitive results.

Findings

Achieved an F1 score of 0.7368, close to the top score of 0.75.

Utilized a small set of semantically-informed features.

Showed effectiveness of knowledge graph-based features in attribute classification.

Abstract

Luminoso participated in the SemEval 2018 task on "Capturing Discriminative Attributes" with a system based on ConceptNet, an open knowledge graph focused on general knowledge. In this paper, we describe how we trained a linear classifier on a small number of semantically-informed features to achieve an $F_1$ score of 0.7368 on the task, close to the task's high score of 0.75.