Trying to reformat my nlu data for rasa 1.10.2 (to run on rasa x)
running this command from the documentation
sed -i -E 's/\[([^)]+)\]\(([^)]+):([^)]+)\)/[\1]{"entity": "\2", "value": "\3"}/g' nlu.md
gives me this error
sed: 1: "s/\[([^)]+)\]\(([^)]+): ...": \2 not defined in the RE
I found a solution to this. Sharing for the benefit of anybody who is facing the same problem with changing the training data format.
Open any text editor (like Atom) that supports regex. To convert training data in [value](entity) format for e.g. from this format [FIRST](class_type:FIRST) to this [FIRST]{"entity": "class_type", "value": "FIRST"} format do this:
In find paste this:
\[([^)]+)\]\(([^):]+)\)
and in replace paste this:
[$1]{"entity": "$2", "value": "$1"}
To convert training data in [value](entity:synonym) format for e.g. from this format [first class](class_type:FIRST) to this [first class]{"entity": "class_type", "value": "FIRST"} format
do this
In find paste this:
\[([^)]+)\]\(([^)]+):([^)]+)\)
and in replace paste this:
[$1]{"entity": "$2", "value": "$3"}
It takes a few minutes for the formatting to complete. Once it’s done check to see if everything is formatted according to the new training data format