If you're using a default Microsoft TTS voice like "David" or "Zira", you should be able to use SSML to override the pronunciation, though this does mean you need SSML tags in each action you want it to apply to.
There is some information on SSML use
here(the "Guide the pronunciation of specific words" section should be relevant).
Once you have the pronunciation you want, you could create a .pls lexicon file, which when referenced will apply it to any occurrence of the word between the SSML tags, rather than having to so manually for each one individually.