The brill tagger in nltk, has errors when tagging the instruction data. For example: "heat until brown", would be tagged as "[[('heat', 'NN'), ('until', 'IN'), ('brown', 'JJ')]]". Most of the verbs in the instruction dataset would be tagged as 'NN'.
So I want to modify the rules in the brill tagger, and to change the tag according to my cooking verb list. For example: if the there is no 'VB' before 'IN', and there is a token before 'IN', and tagged as 'NN' in my cooking verb list, then change 'NN' to 'VB'.
However, I don't know how to insert my own rules in the brill tagger (specifically, I am looking at the nltk source code: https://www.nltk.org/_modules/nltk/tbl/template.html). Could anyone help with this?
Thanks!