How to determine the probability of sample for validation in perplexity and what is the differences between documentsTrain and documentsValidation?

1. I have read the code here https://www.mathworks.com/help/textanalytics/ug/choose-number-of-topics-for-LDA-model.html, i confused how to determine the probability of document for validation, in that site, i read the author have choosen 10% from the documents.
2. there are 2 functions: documentsTrain and documentsValidation, before it, there is a function described cvp as 10% of the document, so, what's the different between them?

Thanks.