can use prodigy to build data of rasa chatbot?

can use prodigy to build data of rasa chatbot? is it helpful?

Sure! You could use Prodigy to label entities or top-level intents that apply to the whole text.

Just had a quick look at the Rasa documentation and the JSON format used for training data actually sounds very similar to the format you get out of Prodigy (which makes sense, because in both cases, it represents character offsets in text):

Entities are specified with a start and an end value, which together make a python style range to apply to the string, e.g. in the example below, with text="show me chinese restaurants" , then text[8:15] == 'chinese' .

So when you annotate an example like {"text": "show me chinese restaurants"} with ner.manual and --label cuisine the data Prodigy creates looks like this:

{"text": "show me chinese restaurants", "spans": [{"start": 8, "end": 15, "label": "cuisine"}]}

...which you should be able to use directly as training data for your Rasa bot :slightly_smiling_face:

1 Like