List of English words phonetics

Question

The list actually doesn't have to be free! I just need something that I can buy once or whatever.

And if possible, but not needed, parts of speech.

Before you say anything, I have googled for this extensively and have looked at over 100 links and found nothing useful, except some pre-written rhyming program.

I am needing a list I can use for a programming project, a txt file, xml, or something similar that's not pre-compiled in an unprogrammatically interperable format.

Any help is greatly assisted. I have already viewed every other thread on stackexchange on which this was asked(and other sites of course), and have found no working or useable links.

Answer 1

The cmudict provides phonetic spellings of a sizable number of American English words. The CELEX database is a similar project; you can select which data you want and download wordlists at WebCelex. Part-of-speech data (word class) is also available in CELEX. The CELEX download interface is somewhat frustrating, but you should only need to use it right once to be able to download what you want.

Answer 2

I found that the UCLA Phonetics Lab (Archive) page has word lists with sound samples and IPA spelling. There are no parts of speech unfortunately.

e͡ɪ     hayed 
o͡u     hoed 
ɔ/ɑ     hawed 

Links to

• English
• Language Index
• Details

They mention that only a fraction of data is online. If this is the right format, I'd contact them and ask to share more.

Answer 3

Anki is a flashcard tool for computers and phones. A cool thing is that you can use the decks for other purposes.

As on example, the deck "500 Word English Pronunciation with IPA and Audio" (link) contains 500 words formatted like this:

Lemma   business
IPA     [ˈbɪznɨs]
Audio   [non-mp3 audio]
Explanation     -[ɨ] is either [ɪ] or [ə]
Tags    

To get these cards in .csv (tab-separated) format, you'll need to install the Anki tool to your PC. Then, load the deck and choose the Export option.

business    [ˈbɪznɨs]   [sound:PerfectP (179).wav]  -[ɨ] is either [ɪ] or [ə]

(Pastebin link to this specific deck)

Answer 4

The CMU Pronunciation Dictionary sounds like it would be useful for something like this:

http://www.speech.cs.cmu.edu/cgi-bin/cmudict

It has pronunciation guides for 134,000 English words and uses the ARPAbet, which is directly mappable to IPA (https://en.wikipedia.org/wiki/Arpabet)

Answer 5

Look at the Wiktionary. It contains phonetics for many words, and it distinguishes part of speech consistently for its lemmas. And it is free data.

Answer 6

You can also checkout these word lists. I know its a Scrabble based site, but they seem to have everything from what I can tell. I use them all the time and they have some nice json exports as well.

Answer 7

Just list words that you are interested in and script phonetic symbols from a dictionary website. (I use ldoce.)

You could get something like this: {'happy': 'ˈhæpi', 'birthday': 'ˈbɜːθdeɪ'}

Answer 8

Here is a word-IPA mapping list converted from cmudict: menelik3/cmudict-ipa, and it also contains the brown-frequency-list-with-ipa.txt based on the most frequent 33862 words by the Brown Corpus.

Answer 9

For those who are looking for a British English one, here is a list of 123312 words with their corresponding Received Pronunciation in IPA: https://gist.github.com/graphemecluster/bd216027d3a9f3f094794725b02c2775

The pronunciations are generated by eSpeak NG with the following command:

awk '1; {print ""}' en-GB.txt > en-GB_LF-doubled.txt
espeak-ng --ipa --tie -q -v en-GB-x-RP -f en-GB_LF-doubled.txt --phonout en-GB_IPA-only.txt
paste en-GB.txt en-GB_IPA-only.txt > en-GB_IPA.tsv

(You may use espeak instead of espeak-ng for legacy eSpeak versions.)

The input consists of only ISO basic Latin alphabet A to Z and a to z with no diacritics (accent marks), with the straight quote ' being the only character outside the alphabet.

The consonants appeared in the output are:

b, d, d͡ʒ, f, h, j, k, l, m, n, p, s, t, t͡ʃ, v, w, x, z, ð, ŋ, ɡ, ɹ, ʃ, ʒ, θ

And the vowels are:

a͡ɪ, a͡ɪ͡ə, a͡ʊ, e͡ə, e͡ɪ, iː, i͡ə, uː, æ, ɐ, ɑː, ɒ, ɔː, ɔ͡ɪ, ə, ə͡ʊ, ɛ, ɜː, ɪ, ʊ, ʌ

and a special one, /ə͡l/, that probably reflects the vocalisation of l. (If you don’t like it, just do a global replacement.)

Some noteworthy points are:

• There are a lot of uncommon and technical words (even “ECMAScript”, “Bugzilla” and “PyTorch”!)
• The original list is intended for autocompletion and autocorrection, so a plenty of words end with 's.
• Syllable boundaries are unmarked.
• Ties are used to indicate that the adjacent characters are of the same phoneme (i.e. consonant or vowel) (and thus the same syllable) and should be considered as a whole.
• The stress marks (ˈ and ˌ) are placed right before vowels. Originally a eSpeak feature, this effectively facilitates rhyme searching (remember this was the original intent of the OP) without needing the list of phonemes in the program.
• There are at least one primary stress mark (ˈ) for every word.
• Spaces may exist at lexeme boundaries in the transcription.
• ɹ and the opentail ɡ are used throughout the transcriptions instead of r and the looptail g.
• Many schwa (/ə/) syllables at the start and all schwa syllables at the end of a word (with the exceptions of “the” and “yer”) becomes /ɐ/.

Some obvious mistakes (which are probably only a small part) are corrected, and phonemes alien to most English speakers, like /ç/, /ɬ/ and nasalised vowels, are manually eliminated.

Disclaimer: Except removing phonemes, the output is subject to minimal modification and checks. Use it at your own risk.

Background: Like the OP, I made this for my programming project about rhyming.

Hint: For those lucky enough to get here who are also writing a rhyming program, again, you don’t need to hard code the list of phonemes into your app! You just need to consider the position of the primary stress mark (ˈ).

Statistics

Statistics of the data can be found at: https://gist.github.com/graphemecluster/e59634a18d1d2927293aa7f40bb060e5