Temporal Event Knowledge Acquisition via Identifying Narratives

TL;DR

This paper introduces a novel weakly supervised method for extracting temporal event knowledge from narratives, leveraging the double temporality characteristic to improve temporal relation classification and narrative understanding.

Contribution

It presents a new approach that exploits narrative structure to acquire temporal event knowledge across sentences, outperforming existing neural models on related tasks.

Findings

Successfully identified 287k narrative paragraphs from large corpora.

Extracted rich temporal event knowledge that enhances temporal relation classification.

Outperformed recent neural models on the narrative cloze task.

Abstract

Inspired by the double temporality characteristic of narrative texts, we propose a novel approach for acquiring rich temporal "before/after" event knowledge across sentences in narrative stories. The double temporality states that a narrative story often describes a sequence of events following the chronological order and therefore, the temporal order of events matches with their textual order. We explored narratology principles and built a weakly supervised approach that identifies 287k narrative paragraphs from three large text corpora. We then extracted rich temporal event knowledge from these narrative paragraphs. Such event knowledge is shown useful to improve temporal relation classification and outperform several recent neural network models on the narrative cloze task.