I am totally new to RASA and I have tried to build a simple bot, which will understand a reminder with date. So far I have provided ~ 60 training data with the following patterns: “Remind me to … tomorrow at 9 PM”. My goal is to identify the datetime entity and also the todo entity. So far I have tried different pipeline settings and non of them worked well.
What I would like to do is to catch the text that will come after “Remind me to” and make it as a todo entity. That text can be anything (even it can be random characters). I would like to setup RASA, so it will treat it as a free text and not try to analyze it. I have tried different pipeline settings and sometimes it worked, sometimes DIET identifies multiple todo entities from one command. For example, using the default pipeline: “Remind me to talk to my parents in the family chat on messenger tomorrow at 9 PM.” >> RASA will identify “talk to my parents” as a todo; “in the family chat” as a todo…, instead of the whole text as a todo. I have tried to fine-tune the pipeline, but no luck.
I have also trained a wit.ai model and they have the following configuration for entities:
Free Text (An entity that does not belong to a predefined list. You’ll use a free-text entity when you’re open to new values.)
Keywords (An entity that belongs to a predefined list.)
Free Text & Keywords (An entity that is defined by a predefined list, but also open to new values.)
I have used Free Text and wit.ai did a good job identify my entities.
Would you be able to help me how to setup my rasa pipeline to achieve the same result as I can with wit.ai using Free Text lookup? I couldn’t figure out how to train a model, which will understand my intent (create a reminder) and extract the datetime entity as well as the todo entity (which is basically the text that comes after “Remind me to”). I would appreciate any help.
Thank you, Endre