April 30th Post

In a previous post regarding essential questions back in November, I presented two questions that guided my research and progress from that point to now. Those two questions were: "How can we create a program that recognizes spoken names and proper nouns that are spoken by anyone?" and "Is it possible for this program to be as accurate as a program that does use personalized pronunciation?" Writing this now I feel confident in my ability to answer these questions. In regards to the first of the two questions, I have since settled on on a phonetic model. I used a series of unique phoneme signatures in tandem with various generally applicable rules and logic along with their exceptions to make this approach viable. I chose this route largely due to the last part of that first question regarding the program being useful to anyone. For this reason I chose a phonetic approach as it simply operates under the assumption that the user is saying names that largely consist of the basic 44 phonemes. Also, knowing that names don't necessarily follow the same rules as words, breaking them down into phonemes still would allow me to employ various other rules regarding phonetic sequences. Considering that this approach utilizes rules rather than machine learning, the very success of it lends itself to be an answer to my second question. In my own trials with the software I developed based on this approach I have found, conclusively, that this program is indeed more accurate than a model based on personalized pronunciation considering its use with various different people.