from pysle import isletool

isletool.LexicalTool('ISLEdict.txt').lookup('cat')

I think u should point all data files into MANIFEST.in file

See https://realpython.com/pypi-publish-python-package/

I do something like for persian language: https://github.com/PasaOpasen/PersianG2P

Say I want to disambiguate two pronunciations of the word "bow". I can get these from your library (small excerpt from a large number of matches):

>>> isleDict.search('boʊ')
... ('bow', [('# b ˈoʊ #', ['nn', 'vb', 'vbp'])]), ('bow-and-quarter_line', [('# b ˈoʊ . n̩ # k w ˈɑ ɹ . t ə ɹ #', [])]),  ...

I'm having trouble figuring out how to modify the regex to only get '# b ˈoʊ #' without the extra stuff. Any suggestions? I tried the regular anchor characters (^, $) without luck.

Hello!

Firstly, thank you so much for your hard work on this. I'm working on a project to try and make reading more accessible, and your ISLE dictionary interface is a huge help. What I'm trying to do is visually tie together the pronunciation with the letters it stems from.

Unfortunately, this is pretty tricky in English, particularly at the phonetic level. I assume it is quite doable at the syllabic level.

I found the MOBY hyphenator project, and it seems I could merge the two dictionaries, however it is inconsistent with ISLE, for instance:

eve¥ry¥day everyday(jj) # ˈɛ̃ . v ɹ i . d ˌei #

In addition, it seems to be missing some very basic words like:

$ grep -E '^ha[\^rd]+en' mhyph.txt | head -n1
hard^en
$ grep -E '^ha[\^rd]+er' mhyph.txt | head -n1
$

Any ideas on how to go about this?

Thanks! Jacob

Let's see:

import pysle

pysle.isletool.LexicalTool.lookup('cat')
Traceback (most recent call last):

  File "<ipython-input-1-09f456e372af>", line 3, in <module>
    pysle.isletool.LexicalTool.lookup('cat')

AttributeError: module 'pysle' has no attribute 'isletool'


from pysle import isle
Traceback (most recent call last):

  File "<ipython-input-2-23a7e2ed01cb>", line 1, in <module>
    from pysle import isle

ImportError: cannot import name 'isle' from 'pysle' (C:\ProgramData\Anaconda3\lib\site-packages\pysle\__init__.py)


from pysle import pronunciationTools
Traceback (most recent call last):

  File "<ipython-input-3-23b867baacbb>", line 1, in <module>
    from pysle import pronunciationTools

ImportError: cannot import name 'pronunciationTools' from 'pysle' (C:\ProgramData\Anaconda3\lib\site-packages\pysle\__init__.py)

the vowel sound 'ɛ̃' is not accounted for in the search functionality of isletool.

This means that a search like: isleDict.search("",numSyllables=2,multiword='no',stressedSyllable='only',wordInitial='only') should include ('welder', ['# w ˈɛ̃ l . d ɚ #']) but it does not (unless I'm misunderstanding something?)

I was able to fix it for my needs by adding 'ɛ̃' to isletool.charList and isletool.vowelList, and changing line 211 of isletool.py to:

"V": u"(?:aʊ|ei|oʊ|ɑɪ|ɔi|[iuæɑɔəɛɪʊʌɛ̃]):?",  # vowels

I don't know enough about linguistic to know if "." is a phone but it comes up for some words. In some lines in my dictionary like "pirouette" the entry ends with ". #" leading to "." showing up as a phone. Is this intended?

from pysle import isletool
isleDict = isletool.LexicalTool('../data/isledict/ISLEdict.txt')
isleDict.data["pirouette"]
#> ['# p ˌɪ . ɹ u . ˈɛ̃ . #', '# p ˌɪ . ɹ u . ˈɛ̃ t #']
isleDict.lookup('pirouette')
#> [([['p', 'ˌɪ'], ['ɹ', 'u'], ['ˈɛ̃', '.']], [2, 0], [0, 1]),
#> ([['p', 'ˌɪ'], ['ɹ', 'u'], ['ˈɛ̃', 't']], [2, 0], [0, 1])]

Right now isledict is read fully into memory. This is probably fine if an application needs to be able to quickly access items over a long period, but for most use cases is probably unnecessary.

I'm wondering if it might be possible to quickly search for desired content and load only that into memory?

sentence = "do you want another pumpkinseed"
phoneList = isletool.transcribe(isleDict, sentence, 'longest')
print(phoneList)

> 'du ju want ənʌðəɹ pʌmpkɪnsid'

I'm not sure if this is intended but multi word entries only return the first words pronunciation. E.g. for "australian_seal" we only get the pronunciation for australian.

from pysle import isletool
isleDict = isletool.LexicalTool('../data/isledict/ISLEdict.txt')

isleDict.data["australian_seal"]
#> ['# ɑ . s t ɹ ˈei l . j n̩ # s i l #']

isleDict.lookup('australian_seal'))
#> [([['ɑ'], ['s', 't', 'ɹ', 'ˈei', 'l'], ['j', 'n̩']], [1], [3])]

Thanks for making this package!

The intention with version 4 is to modernize pysle a bit

• [x] lazy load isle dictionary for faster boot times (~20s -> 0.5s)
• [x] type hinting
• [x] move to object oriented
• [x] function signature/return type documentation
• [x] unit tests
• [x] simple use tutorial

This PR does the following:

• [x] fixes several important bugs related to syllable alignment
  • a variable set inside a loop was never properly initialized (if there is not stress specified by ISLE, a stressed syllable of the current word would be marked the same as the previous word)
  • a argument to a function was mutated inside of the function (if there were multiple entries for a word in ISLE, later entries would be penalized stronger compared with earlier entries)
• [x] adds more flexibility in error handling with praattools.syllabifyTextgrid
• [x] requires praatio 5.0 or greater
• [x] drops support for python 2.7

Phones in the ISLE dictionary are in IPA. There is no constraint about what the input to ISLE is. pysle tries to smudge the input to be ipa-like but this is kindof hidden to the user. Non-IPA representations probably work sometimes but this is a dangerous assumption.

It's also possible that some parts of the code forgot this fact and assume the input is NOT IPA.

The input representation /is/ preserved in the output, so that is good at least.

Seems like a headache to validate everything is working ok.