grzegorz 0.4.2

A minimal pair generator

grzegorz

grzegorz - a minimal pair generator. For a detailed history of the project, check the Changelog

If you already know about minimal pairs, or are just interested in linguistics, you can use this project as learning material.

If use Anki to learn languages, you can use grzegorz to make an Anki deck full of minimal pairs, helping you learn the sounds of your target language faster. Check the Usage section!

But if you're not sure what minimal pairs are and what they're used for:

Consider that you're learning a new (real) language: like any beginner, you want to learn the phonetics, both to be able to hear words and to be able to speak (Tip: if you learn to it at the beginning of your journey, you won't have to struggle fixing bad pronunciation habits later on). But, in order to learn the sounds, you have to be able to differentiate between them. You won't get far if you keep messing up your ous and us in French.

Enter: minimal pairs - pairs of words that differ by only one sound. Think: bit - pit (english), rue - roux (french) etc. Or, more abstractly, when you test yourself, a sound is played and then you have to choose between two very similarly spelled words. After a few tests, your ability to know what sound you heard increases, even outside of said tests.

Getting started

Requirements

For building:

• python3
• pip

Installation

Make sure the pip installation directory (default: ${HOME}/.local/bin) is on your ${PATH}. If not, add it (to your .bashrc or .zshrc preferably):

export PATH="${HOME}/.local/bin:${PATH}"

From PyPi

pip install grzegorz

From source

Clone this repository and run pip:

git clone https://github.com/xylous/grzegorz grzegorz
cd grzegorz
pip install .

Usage

If you want an to get an Anki deck full of minimal pairs without having to bother too much with boring details, use fullmake:

grzegorz fullmake <language> <numwords> [--clean]

In other words, you tell it what language you want your minimal pairs in, and the number of words to sample (the higher the sample, the more possible minimal pairs found, although the runtime is longer), and, optionally, if it should remove the files it made when running (--clean option).

So if you wanted to sample the 5000 most common Polish words (like me), you'd do:

grzegorz fullmake "Polish" 5000

Check the "Importing into Anki" section for information on how you can use the deck in Anki-proper.

But if you want to manually go through the processes of creating a minimal pair deck, then read ahead.

Manual

There are four commands, each corresponding to a single stage in the creation process:

• wordlist <language> <numwords> <output-file>. The source for the wordlist is a frequency list based on movie subtitle occurences (check hermitdave's FrequencyWords repository). Please don't edit the wordlist file, as it will most likely result in errors.

• fetchipa <wordlist> <output-file>, which takes the output of wordlist and creates a JSON file where all words are paired with their International Phonetic Alphabet spelling, which is fetched from Wiktionary.

• generate <ipa-json> <output-file> [--no-optimise | --ignore-stress], which takes the JSON file created by fetchipa, and outputs the list of minimal pairs it found, in JSON format as well.

  Note that, by default, it's optimised: it filters out pairs with "boring" differences which are easy to tell apart by most people ('q' and 't', 't' and 'd', 'e' and 'o' etc.). Give it the --no-optimise option to not curate the list.

  Secondly, syllable stress marks (., ˌ, ˈ) are kept. You can use the --ignore-stress flag to discard them when generating minimal pairs.

• makedeck <minpairs-list>, which takes the output file of generate and creates an Anki deck package (./grzegorz-anki-deck.apkg) which you can import from inside Anki

Concrete example

Let's assume you want to make a deck full of minimal pairs in French. and let's assume you didn't leave the installation directory (or else python won't find the grzegorz module):

grzegorz wordlist "french" 5000 wordlist.txt
grzegorz fetchipa french-wordlist.txt french-ipas.json
grzegorz generate french-ipas.json minpairs.json --ignore-stress

If you were to specify the wrong wordlist language, shame on you: you'll likely end up with the wrong International Phonetic Alphabet spellings or, worse, none at all.

Then you could generate an Anki deck (output file: grzegorz-anki-deck.apkg, in the current directory, no matter where the input file is located):

grzegorz makedeck minpairs.txt

Importing into Anki

NOTE: the flashcards made don't have any audio, not because of a lack of interest, but because of a lack of free (as in beer) APIs or libraries that can (legally) furnish audio pronunciations. Forvo's API is paid! You must therefore add it yourself, but it won't take a lot of time.

After you've created a deck package, open Anki, click on Files in the top left corner, then Import, and then find the file and load it. A new deck, "grzegorz's minimal pairs", should pop up. You can rename it, you can move it, you can do anything with it. You also have control over its options, i.e. how often you review it and whatnot.

Roadmap

• fetch a wordlist of most used words in a given language
• fetch the International Phonetic Alphabet spelling for a given wordlist
• generate minimal pairs
  • optimise: look for interesting differences
  • optimise: ignore stressed syllables
  • look for phoneme contrasts
  • look for chroneme differences
  • look for syllable stress differences
• Anki integration
  • create Anki flashcards from the generated minimal pairs
  • export a deck containing the created flashcards
  • add audio pronunciations for every flashcard

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

But, honestly, the greatest contribution you can make is to add International Phonetic Alphabet (IPA) spellings to words on the English Wiktionary, which is the source for all the spellings that grzegorz uses. The more IPA spellings there are, the higher the number of words that can be used, the more possibilities for minimal pairs.

License

GPLv3