I have created a rule based English audio/text to Sri Lankan Sign language model using NLP techniques such as dependency parsing, pos tagging and syntactical reordering.
To evaluate my model, I thought of using WER and NIST metrics. Is the below equation valid for obtaining the NIST value from WERs for each sentence?
NIST =( PRODUCT(1-WER) ^1/N) * Scaling Factor
Where N is the total number of transcription pairs in the test data, 1-W is the word accuracy rate and the scaling factor is given by NIST for the number of sentence sizes