RASA (2.4.0) user input filling up more slot than it should

Rasa Open Source
1

In my RASA AI, one user input was supposed to fill up one slot, but as you can see in the below, the user input ‘mac OS’ filled up both the ‘operating system’ and ‘programing language’ slots. Why is it doing this and how can I fix it?

the result I want

What's your operation system?
operation system：None
programing language：None
Your input ->  mac OS                                                          
operation system：mac OS
programing language：None
What's your programing language?
Your input ->  c
operation system：mac OS
programing language：c

the output of rasa shell

What's your operation system?
operation system：None
programing language：None
Your input ->  mac OS                                                           
operation system：mac OS
programing language：c

nlu.yml (lookup table)

- lookup: os
  examples: |
   - mac_OS
   - windows
   - linux
- lookup: pl
  examples: |
   - python
   - java
   - c

domain.yml

entities:
  - os
  - pl

slots:
  os:
    type: text
  pl:
    type: text

stories.yml

- story: welcome
  steps:
  - intent: start
  - action: fill_slot
  - action: utter_check_message
  - checkpoint: reply_fill_slot

### operation system slot_filling
- story: fill_slot_os
  steps:
  - checkpoint: reply_fill_slot
  - intent: say_os
    entities:
    - os
  - action: utter_check_message
  - action: fill_slot
  - checkpoint: reply_fill_slot  

### programing language slot_filling
- story: fill_slot_pl
  steps:
  - checkpoint: reply_fill_slot
  - intent: say_pl
    entities:
    - pl
  - action: utter_check_message
  - action: fill_slot
  - checkpoint: reply_fill_slot
2

Hi @angel08012000 could you please run rasa shell --debug and then post the full debug log to understand why both slots are being filled?

It would also be useful to see your full nlu.yml file with intents? Are you using entity annotations in your intent examples? That could be handy. Also you could experiment with slot_was_set event in your stories to not rely on specific entity names.