How to Train Your AI

NLU

NLU stands for Natural Language Understanding. Your chat experiences have a free text field in which customer can type their requests and comments. With NLU, you can create so called "intents" concerning a certain topic (e.g. goodbye). Now you enter 10-20 example sentences ("Utterances") related to the topic of goodbye that you think could be a possible input from a customer (e.g. "see you later","goodbye", "bye"). By training the AI, our algorithm generalises your example sentences and is now able to identify most related text inputs on the topic. Then you define a specific response for your intent (e.g. "Have a good day!") to be displayed once a customer addresses this specific intent. An intent can be created by creating an article. In the following, you can find the corresponding fields.

Example:

Intent (Name)

Brand lover

Utterances (Synonyms) - Please use a minimum of 10 possible entries and separate them with line breaks.

I love it
I like your brand
love brand
admire your brand
I like.

Response (Article)

We understand. We love our brand as well! :-)

3 things to keep in mind when training your AI

1. Less Is More

Few, but well-defined intents

2. Feed Well and Sensibly

Min. 10-20 utterances per intent

3. One Word Is Not Enough

Single words should only be entered in exceptional cases