Three focused tools for evaluating translation quality: preprocess and align first, then run neural similarity scoring and LLM rubric review.
Strong CJK-aware sentence segmentation, plus an interactive alignment table to split, merge and sync source, reference and candidate texts.
Neural semantic similarity (multilingual-e5-small). Runs locally in the browser. Document or per-sentence mode with a weak-segment heatmap.
LLM rubric scoring (NAATI/ATA/MQM), qualitative commentary and bilingual-glossary adherence. Uploads glossaries from Excel or CSV.