Comprehend.detectDominantLanguage text length & Confidence

Question

Hello,

How long should be a correct text so that AWS SDK Comprehend.detectDominantLanguage to surely detect its language (minimum number of words)  ?

The use case is the detected text fragments from travel pictures. How is Comprehend language detection influenced by the fact that some words from picture, and thus from phrase, might not be correctly extracted ? Will Comprehend guess the misspelled words, or ignore them ?

The answer should lead to a pre-process of the text, so that only high chances text to be sent to Comprehend. How should input text be pre-processed (classified) considering the detection Confidence from Rekognition ? Which is the average Rekognition.Confidence of words for a whole text, so that to be successfully detected by Comprehend.detectDominantLanguage ?

Thank you,  
Mihai

Accepted Answer

The API's minimum requirement for a request is 20 characters.
https://docs.aws.amazon.com/comprehend/latest/APIReference/API_DetectDominantLanguage.html

For questions about accuracy and integration with other services (Amazon Rekognition), please cut a ticket to be investigated with the necessary resources.

Comprehend.detectDominantLanguage text length & Confidence

Relevant content