SpeechRecognizer filtering filler words

Hello! I'm currently trying to create an app which counts the amount of times you say a specific word. I'm orienting it towards filler words such as "um", "hm", "uh", "ah", but it's intentionally filtering out those words, which defeats the whole purpose of my idea. Is there any way to have it not do that? Extensions are fine if there is no way to modify the object itself.

Here is the project:
Um.aia (114.7 KB)