{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/be7b5804b9e849a182981b88e66141ec\" frameborder=\"0\" width=\"1334\" height=\"1000\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1000,"width":1334,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1000,"thumbnail_width":1334,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/be7b5804b9e849a182981b88e66141ec-ce7612a401e64f65.gif","duration":373.572,"title":"ContextMinder: Context-Aware Localization QA Tool","description":"This Loom presents ContextMinder, a context-aware localization QA tool that combines deterministic checks with OpenAI LLM validation. The speaker demonstrates it using a made-up 19th-century Poland RPG dataset to test Polish and German translations, including style guides and character descriptions. They report sending 85,000 tokens and receiving almost 2,000 tokens, and then show detected issues such as feminine-form misuse in Polish for a male character, missing or incorrect tags, grammar and capitalization errors, and an incorrect button translation. Examples also include LLM suggestions to refine strong wording like Cuchniesz and verify register-appropriate choices."}