Alternatives for load_conll

[Dashka Rodionova]

Hello,

I use your toolkit in sequence learning. I both train and predict data with csv/tsv files.
In the other hand, it would be useful to make predictions immediately from code (i.e, POS-tagging for a sentence as function argument, not for a sentence from file).
Does seqlearn get data only from files? Is it possible to implement an alternative for Pandas dataframes loading?

Thanks,
Daria