How to fill slots with any word using nlu.md

Rasa Open Source
1

Hi,

I want to better understand how slots are being filled and how can i force it using the nlu.md file. For example i want to create a simple bot where i can tell him to remember something (can be any word), use a slot to store that word. When i come back and ask for that slot name, he should be able to tell me what value it points to.

So far he never stored the value in that slot except is the word written in brackets, eg: “word”, “beautiful”

For example i tried these options: nlu.md

## intent:remember
- remember this info [something](slot_to_remember)
- store this in memory [word](slot_to_remember)
- please remember this [beautiful](slot_to_remember

## intent:display_mem
- what you have in slot_to_remember?
- display slot_to_remember
- what is in slot_to_remember?

**stories.md**
## remember_this
* remember
  - utter_affirm

## memory_test
* display_mem
  - utter_from_memory

**domain.yaml**
intents:
 - remember
 - display_mem

slots:
  - slot_to_remember
      type: text

responses:
  utter_affirm:
  - text: "ok"
  utter_from_memory:
  - text: "You told me {slot_to_remember}"

Using the above example, how can i have him correctly store the following words in these senteces:

  • remember this banana
  • please remember this computer
3

I am already investigating different ways to retrieve that information, thank you. However my problem is storing the information into slot_to_remember. I know how to do it with entities, however thats not what i need. For example today i may tell him to remember current time, and tomorrow a brand of a car. he should be able to do so as long as i use the same construct over and over again:

intent:remember

  • remember this info something
  • store this in memory word
  • please remember this [beautiful](slot_to_remember
4

hi @remenyic - it seems to me that the problem is not enough training data. If you want the model to learn to recognize this entity well (including never before seen values) you will need more training examples.

Of course, if you only want it to work for the three phrases “remember this”, “stores this in memory” etc, then you could just use a regex and don’t need an NLU model

5

Ah, it looks like i choose the harder path. Thank you, that clarified it for me.