Prefer names when homophone encountered #3736
Description
My name is Christi and I have a room "Christi's room" and entities like "Christi's mood light". I expect Assist to find these, but it does not. On Chrome browser on my macbook, it usually spells the name as "Christie". On my iPhone, it usually spells it "Christy". It has never spelled it the way I spell my name. There are also many alternate spellings that I have not seen such as "Kristy".
When Assist uses the incorrect spellings of my name, it cannot find the devices or area. In the following screenshot, the first two requests were parsed from my microphone while I typed the last one.
The same problem likely occurs for other homonym names like Greg/Gregg, Chris/Kris/Criss, Hailey/Hailee/Haleigh/Haley/Haylee/Hayleigh/Hayley/Haylie, Aaron/Erin, etc. I'd expect the same issue to occur for areas and floors with homonym nouns although I can't think of any real world examples that don't use a person's name.
I expected because my name is included in the area and entity names, that the parsed sentence would spell my name correctly and find the correct devices or room. Based on looking through prior issues and the code, I was wondering if the apostrophe was stripped out and caused the word to become "Christis" so I exposed person.christi to Assist but that did not help. (side note: the response is nonsense and I'm not sure what it thought I was asking, but I expected the answer to be yes because I am shown as home when displayed as a roaming entity on the dashboard.)
As a workaround, I've added aliases for the spellings it has heard, but adding multiple aliases for every entity and room is tedious, and I believe it should work the way I expected. I am surprised that I was unable to find other reports of this issue so I may be in the wrong place, please let me know! I did find issue #1493 which asks for homonym support for French and German but those are specific common words, and was hack-fixed by adding a list in yaml, so a similar approach would not work for proper nouns since there are so many. Perhaps it could work if we were able to define our own list in our personal configuration, but I didn't see a way to do that, and I don't think that's likely to work with cloud parsing.