No Related Conversation if no Tokens are extracted

[westei]

Currently no Related Conversations are suggested if no Tokens are extracted for a Conversation.

As the Conversation MLT implementation does not require any extracted Tokens this can be considered as Bug

[ruKurz]

How to test this issue?

[westei]

You need to create an Conversation with a message where no Token is extracted. In setups without InterestingTerms this is rather easy as you just need to avoid the mention of Named Entities like Person, Organisations and Places.

In a Setup with Interesting Terms its much harder to create Conversations where no Tokens are extracted. What should work is to use Messages that mention Words that do not exist, but still sound German to not fool Language Detection

I used a setup without Interesting Terms to test this more easily, but a sentence like

Was ist ein Verhausenbingler und wozu braucht man das?

is identified as German and results in no tokens to be extracted even in a setting including Interesting terms.

If you do not get related conversations you might need to close the first conversation and create a 2nd one mentioning "Verhausenbingler".

Ich habe den Verhausenbingler noch immer nicht verstanden.

now you can expect getting the first conversation based on the term "Verhausenbingler"