Tackling Hallucinations in Neural Chart Summarization

Saad Obaid ul Islam, Iza Škrjanec, Ondřej Dušek, Vera Demberg
8/1/2023
cs.CLcs.LG

No official code found. Best implementation: WorldHellow/Hallucinations-C2T (2 stars, last commit 2024-10-26).

Abstract

Hallucinations in text generation occur when the system produces text that is not grounded in the input. In this work, we tackle the problem of hallucinations in neural chart summarization. Our analysis shows that the target side of chart summarization training datasets often contains additional information, leading to hallucinations. We propose a natural language inference (NLI) based method to preprocess the training data and show through human evaluation that our method significantly reduces hallucinations. We also found that shortening long-distance dependencies in the input sequence and adding chart-related information like title and legends improves the overall performance.

View on arXivView PDF

Code Implementations(1)

WorldHellow/Hallucinations-C2TCommunity100%
20Jupyter NotebookJan 16, 20231 years ago

Similar Papers

Correction with Backtracking Reduces Hallucination in Summarization

89% match1 repo12023

Mutual Information Alleviates Hallucinations in Abstractive Summarization

85% match1 repo132022

Improving Factual Consistency of News Summarization by Contrastive Preference Optimization

85% match1 repo72023

ChartInsighter: An Approach for Mitigating Hallucination in Time-series Chart Summary Generation with A Benchmark Dataset

85% match1 repo102025

Critic-Driven Decoding for Mitigating Hallucinations in Data-to-text Generation

84% match1 repo72023

Building Reliable Long-Form Generation via Hallucination Rejection Sampling

84% match1 repo12026

Cite this paper

@article{islam2023tackling,
  title  = {Tackling Hallucinations in Neural Chart Summarization},
  author = {Saad Obaid ul Islam and Iza Škrjanec and Ondřej Dušek and Vera Demberg},
  year   = {2023},
  eprint = {2308.00399},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2308.00399v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2308.00399.svg)](https://codeofpaper.com/paper/2308.00399)

Discussion