Rasa Algorithm Whiteboard - TED in Practice

[@jayhu6075]

Thank to you great explanation it make my life easier about this topic. I have question, which pipeline tokenizer, featurizer settings must i use for the Dutch language?

[@RasaHQ]

This depends on your dataset and this is something you'll want to varify yourself. You can try the spaCy embeddings in which case you'll need to user the spaCy tokenizer but it'd be good if you started with the example that we list over here; rasa.com/docs/rasa/nlu/choosing-a-pipeline/#the-short-answer

[@jayhu6075]

@@RasaHQ Thx for the info.

[@adidas1793]

I don't understand how user message with multiple actions as response goes in as different inputs to the transformer? How is the concatenated f(t) created for them?

[@theunknown2090]

So adding more stories should fix this problem right ?

[@rahulmukerjee477]

Can the bot_challenge be inserted differently like so: * confirm - utter_x_countdown * bot_challenge - utter_iamabot * confirm - utter_x-1_countdown ... Is there a reason to test it the way it is shown in the video instead of the above? Thanks.

[@RasaHQ]

(Vincent here) Ones caveat first: ever since Rasa 2.x we've switched to the `.yml` file standard for our stories. Next, the `utter_x_countdown` is a merely a name of an action. The name does not include any logic, that would need to be handled else-where. You *could* use slots for this though.

[@rahulmukerjee477]

@@RasaHQ Thank you for such a quick reply. All your videos are helpful for all skill levels. And thanks for the clarification about the new syntax To clarify my question, will the model learn differently in the 2 (somewhat similar) cases below, all else being same: *confirm -utter_2 *bot_challenge -utter_iamabot -utter_1 *confirm -utter_2 *bot_challenge -utter_iamabot *confirm -utter_1

[@anamontalvo9449]

it was great! thanks