com.owoga.phonetics — com.owoga/phonetics 0.1.3

cmu-lexicon clj

The CMULexicon can get phones for words that aren't in the CMU Pronouncing Dictionary. But the phones are slightly different. The AH sound, as in allow, is returned as ax from the CMULexicon. Also, unstressed vowels don't have a 0 suffix. Instead, the CMULexicon just returns unstressed vowels as the vowel itself with no suffix.

The above is important to note if you want clean interplay between these two different ways of getting phonemes.

cmu-lexicon->cmu-pronouncing-dict clj

(cmu-lexicon->cmu-pronouncing-dict phonemes)

The CMULexicon returns the AH sound, as in allow, as ax. The Sphinx dictionary treates that sound as AH. This converts ax to AH. It also adds 0 to phonemes that are unstressed, which CMULexicon returns as the plain phoneme with no stress marker.

cmu-word-alternatives clj

For words with multiple pronunciations in the CMU dictionary, this maps from the word to its variations. reputed -> reputed, reputed(1), reputed(2).

Not particularly useful itself since reputed(1) doesn't tell you how it's different from reputed. But it's useful to look up the pronunciations in the CMU dictionary.

cmu-word-to-stressed-phones-map clj

Map of lowercase English words to their phonetic sounding based on the CMU Pronouncing Dictionary at http://www.speech.cs.cmu.edu/cgi-bin/cmudict/

Includes words with apostrophes, like possessive aaronson's.

Words with multiple pronunciations have keys with a (1) or (2) after their duplicates, like [aaronsons(1) (AA1 R AH0 N S AH0 N Z)]

Primary stress is indicated by a 1 after the phoneme. Secondary stress with a 2. Unstressed with a 0.

cmu-word-to-unstressed-phones-map clj

consonant clj

get-phones clj

(get-phones word)

Tries to get phones first from the CMU Pronouncing Dictionary and falls back to the CMULexicon if the word doesn't exist in the dictionary.

Input must be lower-case.

Returns a vector of all possible pronunciations.

get-word clj

(get-word phones)

long-vowel clj

phonemap clj

remove-stress clj

(remove-stress phonemes)

short-vowel clj

single-sound-bigram clj

stressed-phones-to-cmu-word-map clj

The same sequence of phones can map to multiple words.

syllable-end clj

unstressed-phones-to-cmu-word-map clj

There might be unstressed phones that can map to two different pronunciations when stress is added, so this maps unstressed phones to a vector of words that can be looked up in the CMU Pronouncing dictionary to see what their stressed phones are.

Another example, look at how many words map to [N IY S]. [[N IY S] [neice neece niece nice kneece kniess neiss neace niess]]

vowel clj

word-alternatives clj

(word-alternatives word)

For words with multiple pronunciations in the CMU dictionary, this maps from the word to its variations. reputed -> reputed, reputed(1), reputed(2).

Not particularly useful itself since reputed(1) doesn't tell you how it's different from reputed. But it's useful to look up the pronunciations in the CMU dictionary.