Direct localization engines for English, French, Chinese, Arabic, and Mongolian, including dialectal negation handling to prevent polarity mismatch.
Generic sentiment models trained mainly on English routinely misread negation, slang, and dialect in other languages — a literal translation can flip a phrase's real meaning. Glimpse runs dedicated engines per language instead of one model translated on the fly.
A brand operating across multiple markets can't afford a reputation score that's silently wrong in a language your team doesn't speak fluently. A confidently-wrong score is worse than an obviously-missing one.
English, French, Chinese, Arabic, and Mongolian today, each with its own dedicated parsing engine rather than one model translated on the fly.
Literal translation can flip a phrase's real meaning — negation, slang, and dialect are easy to get backwards. Glimpse parses each language directly, including dialect-specific negation handling, instead of scoring a translated version.
The dialect- and negation-aware parsing is specifically designed to catch cases where a literal reading would misclassify sentiment — for example, an Arabic clitic negation construction that flips a positive root word to negative.
The current confirmed set is English, French, Chinese, Arabic, and Mongolian. Any expansion beyond that would need its own dedicated parsing engine, consistent with how the existing five were built.