Different `spell_corrected_text` and `correction_dict`?

[varundey]

Hi, thanks for building this. I am seeing some weird results with the output of the model. I have trained the model on 49k rows using term frequency and correcting icea creamq outputs spell_corrected_text same as the original input but in correction_dict I can see that model has identified the correct spelling of the term as ice cream.

What is the correct behavior as the model and as the user of the model?