So we're having an issue where it seems one or two keywords are too heavily weighted in our Messaging bot article recommendations. 

In our company's case, the word "Design" is used quite frequently for a variety of  intents throughout our help centre articles. 

The problem we're having is the article recommendations are almost always never what the customer is looking for because the main keyword the messaging bot seems to pick up on is "Design" rather than everything else in the customers question. 

For example, we have a large section of our help center devoted to accessibility and how users can make their designs accessible. If a user asks the question "How do I make my design accessible" No accessibility related articles. Instead numerous completely unrelated articles are provided simply because the word "Design" is included in them. 

I guess what this boils down to is there doesn't seem to be a way for me to tell the messaging bot specifically to ignore common words, in our case, the word "design". 

And yes I realize the best option here is to craft an answer for the bot if I know I want to target the accessible intent. The problem surfaces for lesser known intents we don't have crafted answers for yet and the negative user experience customers receive when chatting with the bot. 

I'll also highlight that this is NOT the case for email article recommendations. The email auto reply suggestions actually handle this MUCH better and customers get the articles that make sense. 

I hope that makes sense? TLDR, we have a broadly used keyword that seems to be too heavily weighted causing completely inaccurate recommendations within messaging.