Corrected versions of Moby lexical database

The original Moby database is in the public domain and is available at https://icon.shef.ac.uk/Moby/

So far, this repository contains only the pronunciation database, with some corrections.

Standardisation

• Change all line endings to Unix standard (single CR, 0x0a)
• Remove trailing white space from all lines
• Replace 'S' with 's' everywhere, there is no difference
• Replace 'W' with 'w' everywhere, there is no difference
• Replace all "a" with "A", signifying IPA "a"
• Replace all "/ju/" with more often used "/j//u/"

Errors

• Replace spaces with underscores inside words or phonetic encodings
• Remove underscores in phonetic encodings of words not containing underscores
• Replace 'c' with 's' or 'k' as necessary
• Replace 'x' with 'gz', '/x/' or 'ks'
• Remove doubled stress markers
• Miscellaneous fixes to bad or incomplete encodings
• Arabic pronunciation of 'Bent' is closer to 'b/I/nt'
• Remove parenthetical comments mistakenly included in phonetic encoding
• Remove some non-ascii characters and replace with best guess
• Replace double slashed '//Oi//' with '/Oi/'
• Fix many wrong usages of bare vowels like "a", "u", "i" with correct phoneme
• Replace all instances of "/oU/r" with "/O/r" (rhymes with "score", "north" etc)

Editorial decisions

• Replace many 'e' in French words with '/E/'
• Replace 'Z' in French words with 'z'
• End 'chutspah' with 'tsp/@/' not 'tzp/@/'
• Replace 'g' in some German words (eg 'Zug') with 'k'
• Replace many 'V' with 'v'
• Encode French ɛ̃ as in "point" as "/E/N" instead of "aN"
• Remove all "O" and replace with appropriate phoneme

Documentation

• Document phonemes used in database but not mentioned in readme
• Explain use of '(...)' as optional phoneme.

Remaining questions

• Should 'Cipriano' begin with 's' or 'tS' (Italian pronunciation)?
• What does /z/ mean? German or Italian sound? If so, it should be 'ts'.
• When should /-/ be used instead of /@/?
• A few other issues and inconsistencies concerning non-English words.
• Inconsistent use of optional /@/ in some words, eg 'operative' and related words.
• What about several symbols used in non-English words which are (a) not described in the original documentation, and (b) are already encoded by existing symbols? Eg, "S" in some French words is just "s", and "/z/" in German is just "ts". Also, many uses of "e" in French words.
• "V" is sometimes Spanish β (like "b") but sometimes Dutch ʋ (like "w")

Other to do

• Reduce redundancy of compound words: find all cases where both parts of a compound word can be correctly encoded independently, ensure the single word entries exist, then delete the compound entry.
• Swap "A" and "a", so that "a" means IPA /a/ and "A" means IPA upside-down a.
• Use "0" instead of "y" and "y" instead of "Y"
• Simplify encoding: use j instead of /j/, x instead of /x/, S instead of /S/, Z instead of /Z/ E instead of /E/, I instead of /I/, D instead of /D/, maybe B instead of Spanish V, get rid of /z/
• Possibly use other sources to cross check: CMU dictionary (included); Oxford Dictionaries API https://developer.oxforddictionaries.com; Lingorado https://lingorado.com/ipa/; Words API https://www.wordsapi.com.