From b3f5668da648a2004f98280a05a1db6bb92ad1c4 Mon Sep 17 00:00:00 2001 From: Sedat Kapanoglu Date: Tue, 14 Jan 2025 12:33:06 -0800 Subject: [PATCH] improve language detection accuracy --- .../com/composer/select-language/SuggestedLanguage.tsx | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/src/view/com/composer/select-language/SuggestedLanguage.tsx b/src/view/com/composer/select-language/SuggestedLanguage.tsx index 472860a321..32513aeca3 100644 --- a/src/view/com/composer/select-language/SuggestedLanguage.tsx +++ b/src/view/com/composer/select-language/SuggestedLanguage.tsx @@ -33,7 +33,14 @@ export function SuggestedLanguage({text}: {text: string}) { const {_} = useLingui() useEffect(() => { - const textTrimmed = text.trim() + let textTrimmed = text.trim() + + // remove the last word before guessing to prevent a half-written word + // from botching the confidence of language detection. + const lastSpace = textTrimmed.lastIndexOf(' ') + if (lastSpace > 0) { + textTrimmed = textTrimmed.slice(0, lastSpace) + } // Don't run the language model on small posts, the results are likely // to be inaccurate anyway.