I have no idea why this behavior occurs in the first place.  it isn’t logical.  As isn’t capitalized and there is no period after it so it isn’t an abbreviation.  It makes no sense for one as to be spoken one way and an identical as following soon after to be spoken a different way.  This sounds like a programming error or bug that, as Brian suggested, would be a good idea to report.

       

As for the speech dictionary, I don’t know how the voice you are using will react, but I just tested my suggestion.  My synthesizer says St. Ives as saint Ives.  If I create a dictionary entry with St. in the pattern field and the letters st in the pronounced as field, St. Ives is pronounced as st Ives.  As I understand the way the NVDA dictionary works, this sort of entry should work regardless of synthesizer.  As I understand it, the screen-reader sends the pronounced as text to the synthesizer so the synthesizer doesn’t even have the chance to know what the original text is that is triggering the unwanted speech. 

 

Gene