When to Use Extra Context: Evidence-Grounded Terminology Adaptation for Simultaneous Speech Translation
2026-07-20 • Computation and Language
Computation and Language
AI summaryⓘ
The authors study how adding extra context helps translate technical talks live, but find that it mainly helps by correctly translating specific technical terms rather than improving overall meaning. They introduce EGTA, a method that creates a small memory of key terms from the document and uses only those terms to improve the translation process without needing to retrain the whole model. When tested on technical speech translation tasks, EGTA consistently improved scores for terminology recall and translation quality. Their tests show the gains come from matching the document’s key terms rather than just adding more general context.
Simultaneous speech translationTechnical terminologyDocument contextASR (Automatic Speech Recognition)BLEU scoreNamed entity recognitionAcronym recallEnd-to-end modelsCascade translationEvidence-grounded adaptation
Authors
Zeyu Yang, Satoshi Nakamura
Abstract
Extra context is valuable for simultaneous speech translation of technical talks, but injecting the entire document context into every streaming segment is often too coarse. Through diagnostic experiments, we find that context gains mainly come from paper-specific terminology recovery rather than uniform semantic enhancement. We therefore propose EGTA, an Evidence-Grounded Terminology Adaptation framework that builds a document terminology memory, selects compact candidate terms conditioned on the current streaming state, and adapts ASR/speech-side and decoder-side decision spaces using only the selected terms. EGTA can be instantiated in cascaded, end-to-end, and generation-only SimulST settings without full-model fine-tuning. We evaluate EGTA on an ACL technical-talk SimulST evaluation suite consisting of MCIF-dev and ACL60/60-dev. On MCIF-dev, EGTA-RG improves BLEU by +1.05/+0.59, XCOMET-XL by +0.019/+0.006, named-entity recall by +79\%/+73\% relative, and acronym recall by +0.099/+0.171 on En$\rightarrow$Zh and En$\rightarrow$De. Across MCIF-dev latency settings, EGTA consistently improves XCOMET-XL, named-entity recall, and acronym recall. External validation on ACL60/60-dev further shows consistent terminology-recall gains without additional fine-tuning. Shuffled-memory controls and activation audits provide evidence that the improvements are tied to paper-specific evidence alignment rather than generic context prompting.