Tackling Hallucinations in Neural Chart Summarization
This work addresses the problem of hallucinations in chart summarization for users relying on accurate data-driven summaries, but it is incremental as it builds on existing methods.
The paper tackled hallucinations in neural chart summarization by identifying that training data often contains extra information, and proposed an NLI-based preprocessing method which significantly reduced hallucinations according to human evaluation.
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.