Confused Behavior about entity extraction form

I have a worried behavior with Rasa test.

We I run rasa test with space between ] { it give me about 90% End-to-End accuracy. Full example in the nlu data is the follows:

- يجب أن أتحقق من [رصيد](amount-of-money) [بطاقة الائتمان] {"entity": "account_type", "value": "ائتمان"} [أولا](account_type)

  • But when I am using it without space ]{ it gave me about 66% End-to-End accuracy. Full example in the nlu data is the follows:

- يجب أن أتحقق من [رصيد](amount-of-money) [بطاقة الائتمان]{"entity": "account_type", "value": "ائتمان"} [أولا](account_type)

Anyone has a solution for this? Or Am I not writing the example well?

@koaning

My first impression is that this be related to text editors trying to “help” you by sorting Arabic. I’ve seen it go awry before, so just to check …

Is this the intended annotation?

يجب أن أتحقق من [رصيد](amount-of-money)

Or is this the intended annotation? (I’m showing a screenshot from sublime text, which doesn’t resort the way that the browser does).

The first one. I am using vsc IDE and it works.

Also, I check the examples on external editor to be sure about the language. The screenshot is the same example you have provided.