Automatic Fact-guided Sentence Modification
This addresses the need for efficient updates in online encyclopedias, offering an incremental improvement in automated text modification.
The paper tackles the problem of automatically updating Wikipedia articles when new information contradicts existing content, proposing a two-step method that achieves the highest SARI score on a fact update dataset and reduces error by 13% when augmenting a fact-checking dataset.
Online encyclopediae like Wikipedia contain large amounts of text that need frequent corrections and updates. The new information may contradict existing content in encyclopediae. In this paper, we focus on rewriting such dynamically changing articles. This is a challenging constrained generation task, as the output must be consistent with the new information and fit into the rest of the existing document. To this end, we propose a two-step solution: (1) We identify and remove the contradicting components in a target text for a given claim, using a neutralizing stance model; (2) We expand the remaining text to be consistent with the given claim, using a novel two-encoder sequence-to-sequence model with copy attention. Applied to a Wikipedia fact update dataset, our method successfully generates updated sentences for new claims, achieving the highest SARI score. Furthermore, we demonstrate that generating synthetic data through such rewritten sentences can successfully augment the FEVER fact-checking training dataset, leading to a relative error reduction of 13%.