-1

I'm trying to make a chatbot in Luis for my college project. I did so much but I stuck.

First I need to give a little pre-info about Turkish prepositions. In the Turkish language, prepositions comes after the word. Example:

"Seattle" is a city name. If we want to say "in Seattle" in Turkish we will say "Seattle'da". "da" means "in".

Now here is my problem: when a chatbot user says "22.12.2020 Seattle'da bla bla." Luis can understand "Seattle" is a city name. But most Turks do not use " 'da ". They write like "22.12.2020 Seattleda bla bla." Without " ' " I mean.

I can train Luis for "da" prepositions but Luis doesn't give me the exact city name. Luis gives "Seattleda" not "Seattle". I need the city name without prepositions.

4

2 回答 2

1

As far as it goes within LUIS the behavior you are decribing is by design that every token is extracted as it is and cannot be separted. To overcome this issue, you have to do post processing yourself on the extracted entity. You need to check if the entity ends with "da" then you remove it before further using the entity in your application.

于 2021-01-04T19:30:44.813 回答
1

What I do in German is to use the normalization feature of list entities. You will get the normalized form back and then it is easy to process it further. But this is realistic only for a closed set of possible words. Otherwise you have to use preprocessing as described by Maged.

于 2021-01-05T16:49:38.220 回答