I used the rules outlined in the 'Benjamin Franklin's phonetic alphabet' Wikipedia article.
Additionally I used the primary source: 'A Reformed Mode of Spelling.' within Franklin's 'Political, Miscellaneous, and Philosophical Pieces'.
Sometimes, the output of this program is inconsistent with examples in the primary source. I have decided that I will stay loyal to the description on the Wikipedia page. This is because (a) I assume that pronunciations have evolved since Franklin's time, and that the Wikipedia authors took this into consideration. This would mean some that the spelling based on new pronunciations causes discrepancies with the old spellings based on old pronunciations. And (b) I don't want this project to devolve into a wild goose chase.
- The Wikipedia article says that eɪ/EY (bait) can "sometimes" be shown with 'e'. It also says "/eɪ/, at the time more probably [eː~ɛː], is represented as ee or e". So I decided to represent eɪ/EY with 'ee' to make 'bet' and 'bait' distinguishable ('bet' and 'beet' respectively).
- The Wikipedia article says that a /z/ at the end of a word is sometimes represented with an 's'. I basically had to ignore this because detecting whether the word should end with 'z' or 's' is out of the scope of this project.
The author of 'A Reformed Mode of Spelling.' says, regarding capital letters, "I should have provided for them by means of larger types, but the form of some of them would have made them too large for the page". So, only old (e.g. 'b', 'i', etc) letters are capitalized in that work, and I have not found the capitalized versions of the new letters. As such, only old letters are capitalized. In addition, 'ʃ' (the 's' of the day) is capitalized when necessary and becomes 'S', and 'ɑ' (the 'a' of the day) is capitalized when necessary and becomes an 'A'.
Franklin's characters are shown here. I primarily used the corresponding Unicode characters from the Wikipedia article mentioned above. But, I wanted each Franklin character to correspond with exactly one Unicode character. So, I used this tool, this tool and this document to find new characters whenever the Wikipedia article suggested using a combination of characters to create a Franklin character.
First, the English words are sent through Carnegie Mellon University's CMUdict to convert them to ARPABET words. If the input word is not in CMUdict, it is simply passed through and outputted in red.
Then, I manually created ARPABETToFranklin.js, which converts ARPABET sounds to Franklin characters. I did this using the content on the 'Benjamin Franklin's phonetic alphabet' Wikipedia article.
The conversion process can be escaped by surrounding words with ampersands. For example, &hello& will be outputted as hello, not hɥlo.