com.apple.SpeechRecognitionCore.speechrecognitiond is the daemon that turns human speech into something the computer can work with, then matches the speech to the words that fit best. As far as I can tell, it gets started and kept running to support three different OS X features: Voice Control, Dictation, and Siri.
Voice Control is a resource hog, just by itself, but if you are not actually dictating with Dictation or actively using Voice Control or Siri, that process should not be using that much of your CPU, and certainly shouldn't be slowing your MBP to the point that you are unhappy with the performance. Typically, this sort of resource hogging happens when there is an actual code bug, but it can happen if the system doesn't get rebooted often enough. When daemons run for long periods, errors can creep in that are not the result of a flaw in the code. Rather, they are usually the result of read/write errors or storage corruption as the code and data get paged back and forth between memory/disk and the CPU itself. Eventually, the errors add up and either cause a runaway process or the process crashes.
I recommend that most people make a point of boot cycling the system at least once a week via -Restart... or -Shutdown. That triggers the housekeeping routines that clear all of the system caches that are prone to causing problems.
It is possible to clear up this particular issue without actually rebooting with the following process:
- Disable Voice Control: ->System Preferences->Accessibility->Voice Control, uncheck "Enable Voice Control."
- Disable Dictation: ->System Preferences->Keyboard->Dictation->Dictation:->Off
- Disable Siri: ->System Preferences->Siri, uncheck "Enable Ask Siri."
- Force Quit the speech recognition daemon: Launch Activity Monitor and select com.apple.SpeechRecognitionCore.speechrecognition.d, then click the "x" button and click the "Force Quit" button in the dialog that pops up.
If you want to, you can now re-enable Siri and/or Dictation without taking any significant performance hit, as those features only load the speech recognition daemon when you using the feature. Voice Control is going to load the deamon and keep it running all the time, though. So, only re-enable Voice Control if you actually use it, because you will have some loss of performance. Also, the daemon should behave properly and die shortly after you toggle Voice Control off again.
Hope this gets you running like you want, helps you recognize when things are not working properly, and lets you make an informed decision about using Voice Control and the other speech recognition dependent features of Catalina.
Cheers!