facebook / duckling

Language, engine, and tooling for expressing, testing, and evaluating composable language rules on input strings.
Other
4.05k stars 719 forks source link

[Persian][Numeral] Extract incorrect number from text #713

Open AmirMohamadBabaee opened 1 year ago

AmirMohamadBabaee commented 1 year ago

when I checked the numeral dimension, I figured out that there is an issue with the numeral extractor. in Persian, صدرا is a first name, and صد means hundred. when the name is fed into the Duckling the output is 100را which is not correct. I tried some other examples and found it in another example when the verb بده (which means give) is converted to ب10 cause ده means ten in Persian. How can I change the configuration of Duckling to enforce rules to apply just to tokens that are space separated?