Pre-processing text for Amazon Comprehend classification

Question

I want to use Amazon comprehend for sentiment analysis. Regarding both the case when using the built-in model or training a custom model of my own is it necessary, or favorable in any way, for me to first do some pre-processing on the text before analyzing it? Such as removing stop words, performing lemmatization etc.

Thanks!

Answer

Pre-processing the text for Amazon Comprehend is definitely not required, I have run Comprehend jobs without any pre-processing and got good results. Whether it is favourable is probably dependent on the specific dataset. I've never tried it and I'd be surprised if the results differed significantly, but the best way is probably to try it yourself with your dataset and compare the results.

Cheers, Heiko

Cheers,
Heiko

Pre-processing text for Amazon Comprehend classification

Relevanter Inhalt