NLU Training

hi all,

I am using the below entities to find an intent

but while testing it works only for “source code” not for “sourcecode”, how do i handle this typo in RASA NLU configuration? And in NLU do i need to capture all words (both singular and plurals) like this (for ex: import and imports), is there way to do in NLU configuration?

Regards, Krishna

Hi @kumarkrish85!

Unfortunately typos are notoriously tricky for classifiers to handle correctly. Within the Rasa toolkit, the best option here is to include all these misspellings, typos etc. in your dataset.

As a second pass, you can try fuzzy matching (though we found it to slow performance down a little). The Rasa developers at CarLabs implemented this in their custom component library Innatis