Effects of Transcript Compression on LLM-based Medical Misinformation Detection in Japanese YouTube Videos

2026-09-25

AI Summary

This study investigates how transcript compression affects the efficacy of large language models (LLMs) in detecting misinformation within Japanese medical videos. By integrating LLMs, RAPTOR-based retrieval-augmented generation (RAG), and J-LIWC linguistic analysis, the research systematically compares classification performance and shifts in linguistic cues across four input strategies: full transcripts, summaries, RAG, and filtered inputs. The findings reveal that transcript compression causes false content to appear more authoritative while diminishing critical detection cues. Full transcripts yield optimal detection performance, whereas all compression methods significantly increase false negative rates. This work provides essential empirical evidence for text preprocessing practices in healthcare misinformation detection.

Abstract

Large language models (LLMs) are increasingly used to assess long-form medical videos, but their effectiveness may depend on whether transcripts are provided in full or compressed through summarization, retrieval, or claim screening. This study examines how such transcript compression affects LLM-based veracity classification of Japanese medical YouTube videos. We compare four transcript input designs: full transcripts, LLM-generated summaries, RAPTOR-based retrievalaugmented generation (RAG), and Screening, which extracts candidate medical and health-related sentences. Using 74 long-form videos labeled as Real or Fake, we evaluate classification performance and analyze linguistic changes using J-LIWC, hedge expressions, and institutional or technical terms. The full-transcript Baseline achieved the best performance, whereas all compressed inputs increased false negatives, meaning that Fake videos were more likely to be misclassified as Real. Summary caused the largest performance drop, while Screening performed best among the compressed inputs but still omitted many medically relevant sentences. Linguistic analyses showed that these errors were not explained by a simple increase in certainty. Instead, Summary reduced affective, social, temporal, cognitive, and conversational cues, while Summary and RAG made institutional and technical terms more salient. These findings suggest that transcript compression can represent Fake videos as more coherent and authoritative inputs, thereby weakening cues needed for misinformation detection

Problem

• medical misinformation detection
• transcript compression
• large language models
• Japanese YouTube videos
• veracity classification

Innovation

• Transcript Compression
• Medical Misinformation Detection
• Retrieval-Augmented Generation
• Large Language Models
• Linguistic Analysis