Incorporating Temporal Information in Entailment Graph Mining

TL;DR

This paper introduces a new unsupervised method to incorporate temporal information into entailment graphs, improving the accuracy of entailment detection in domains with temporally distinct events.

Contribution

It presents a novel approach for injecting temporality into entailment graphs, specifically addressing spurious entailments caused by similar events at different times.

Findings

Incorporating time intervals improves entailment accuracy.

Temporal windowing effectively reduces false entailments.

Model performs well on a manually constructed sports dataset.

Abstract

We present a novel method for injecting temporality into entailment graphs to address the problem of spurious entailments, which may arise from similar but temporally distinct events involving the same pair of entities. We focus on the sports domain in which the same pairs of teams play on different occasions, with different outcomes. We present an unsupervised model that aims to learn entailments such as win/lose $\rightarrow$ play, while avoiding the pitfall of learning non-entailments such as win $\not\rightarrow$ lose. We evaluate our model on a manually constructed dataset, showing that incorporating time intervals and applying a temporal window around them, are effective strategies.