Ist there an order of importance in parameters in machine learning?

I’m doing the course “introduction to machine learning” (

There is an exercise in which you are asked to type in a random sentence and see if the machine learning script correctly processes the content of the sentence as generally positive or negative.

I typed in several sentences beginning with “This exercise is…”. I got a “the sentence is negative as a response” every time. I wanted to test the programme so I initially started out with (what I, a non-native speaker think are) rarer or longer adjectives. So I tried standard ones, such as “good”, “nice” etc… I still got a “the sentence is negative” every single time until I realised the problem was the word “exercise”.

For some reason the word “exercise” is processed as negative and seems to overwrite/weigh more heavily than any adjective given, which surprised me because I thought that nouns might be less powerful when it comes to assessing characteristics than nouns.

Any thoughts on that by a more experienced person in this than I am?

Thanks for your input.

It says in the lesson:
“This Naive Bayes classifier won’t always get the sentiment correct!”

Seems like the word “exercise” is classified/labeled as a verb w/negative connotations.