paralogical / glish

map all words to single-syllable version

Glish

Watch the video about this project: https://youtu.be/sRbcw2sGkJw

Interactive Demo Translator tool here: https://paralogical.dev/glish/

Goal: Make a version of English where every word is only one syllable

Inputs:

• words by frequency (optimize monosyllabification for more common words) inputs/word_frequency.txt
• words with pronunciations and split by syllables (CMU Dict syllablized) Note: multiple valid pronunciations for any given word, but all American english

Stages:

• syllablize.ts → convert CMU dict to JSON mapping of word → IPA split by syllables
• main.ts → load IPA syllables and generate new monosyllabic version of all words
• sonorityGraph.ts → data structure that helps generate new syllables following sonority sequencing.
• respellIPA.ts → convert IPA back into "readable" latin alphabet.

To run code to generate Glish language mapping,

• ts-node syllablize.ts to generate outputs/syllablizedIPA.json + syllableGraph + big list of randomly generated syllables
• ts-node main.ts to generate outputs/monosyllabic.json & other monosyllabic results

To run UI,

• cd ui
• npm run dev

glish's Issues

Excessive spaces when using "Copy monosyllabic" button

"The quick brown fox jumps over the lazy dog" is an English language sentence that contains all the letters of the alphabet. The phrase is commonly used for touch typing practice, testing typewriters and computer keyboards, displaying examples of fonts, and other applications involving text where the use of all letters in the alphabet is desired.

... becomes ...

"The quick brown fox jumps vohdh the bleelz dog" is an glish gwuj stes that ruktntz all the legtz of the feft. The phrase is ahmk used for touch geyept strahrs, tengst preyept and turkst krawrjdz, spahrngspst plizmdz of fonts, and udhd klarnz vlahrng text where the use of all legtz in the feft is deyerdzdzd.

Firefox 120.0.1 on Linux.

antidisestablishmentarianism

The video gave this word as an example, but doesn't have a Glish word for it.

"also" translates to "law"

No need to explain what's wrong with that, I guess :P

pneumonoultramicroscopicsilicovolcanoconiosis is unknown

"pneumonoultramicroscopicsilicovolcanoconiosis" is a famous word online for being long lol, so I guess the glish translator should recognize it?

also multiple weird stuff:
"bookstore" is awrftsk?
"syllable" is libs, which is already a word, and is already an internet slang word for library (mostly used by devs)
"title" is fleyetst, that sounds like german not english 😅
"issue" is shigsht
a coupla more weird stuff that ill post later

most weird of all:
"sonority" is red, so it's unknown?

Show output as IPA?

I struggle with the pronunciation of Glish especially with long words like pawrpstst or stweyedhd but also with words that have an ambiguous pronunciation for my written-english-to-sounds-translator in my head like zawng or veevd.

While it probably wouldn't make me able to pronounce pawrpstst, it would massively help if I had a way to see the pronunciation in an unambiguous alphabet like IPA.

Furthermore, from the video:

Frankly, I think this [IPA] is what the alphabet really should be.

Use monosyllabic slang for some words instead of the generated ones

Some English words already have slang monosyllabic variants:

i.e.

Totally -> Totes
Compensate/compensation ->Comp
Disgust -> Ick

It would be good to research and compile some such words to use in Glish instead of the generated ones.

Glish has some new homophones

One I found is "also law" maps to "law law"

"sonority" is an unknown word

Insane you missed that xD

Ironically, monosyllabic is not a recognized word

As you can see below, despite the word "monosyllabic" being used in the video, on the site itself, and in most dictionaries I can find, the translator does not recognize it as a word.

How many syllables is "diktsksk" (from kiddy as in skript-kiddies)?

I'm already struggling with pronouncing Russian Borshch or the English Scrunched, but Diktsksk (from kiddy) is a new challenge!

Is it possible to add other languages?

Nice project!

Could you provide some guidance, for extending this approach for other languages?

I want to add support of my native language. What kind of data should I collect and how should it be processed to get it working? If it's not too hard, I would love to contribute.

The glish translator only translates the first half of a word containing hyphens, such as "multi-directional".

There's a bug in your application. If you enter a word such as multi-directional, the second word doesn't get translated at all, compared to when it was separated with spaces. I suggest you'd use a tokeniser to pass on non-word characters unchanged and capture groups of word characters, instead of whatever you did to capture only the first half of a hyphenised compound word ;)

Is it necessary to translate names?

Peter:

Lily:

Lisa:

Is it necessary to translate them?
If necessary, why the first letter is in lowercase instead of uppercase?
🤔🤔

Missing word: Uproarious

Missing Word: Pneumonoultramicroscopicsilicovolcanoconiosis

Yes, it was the first word I tried.

Feature Req: Glish to English?

would be cool

Some Glish words clash with existing monosyllable word

For example, the word floor and kink exist as a real, monosyllable words. Revolution and Kinda are non-monosyllable words that got truncated to aforementioned two words.

image

Suggestion: Add a way for users to generate new Glish for unknown words

If we type an unknown word in the Glish translator, then clicking the red word in the translated area generates Glish for that word using the 4 steps described in the video and saves it.

However we should also check that it is indeed an English word:

Step 0) Is it English? (Try finding the word in an English dictionary) -> Return if false
Step 1) Is it already monosyllable? Then save it
Step 2) Try Generating from a syllable, then save it
Step 3) Try Generating from any syllable, then save it
Step 4) Try Generating from any sounds, then save it -> If this fails, then you have hit the limit of all available sound combinations.

Numbers are stripped from the converted text

Example

Input: "Hello, 1, 2, 3, world"
Expected Output: "lhuh, 1, 2, 3, world"
Actual Output: "lhuh, world"

I think it would better if numbers were left alone in the converted text, especially when it comes to copying the text.

"gwitstst" does not seem to me to be a single syllable

I put in "Antigone" as the prompt, and it gave "gwitstst" as the Glish. I'm not sure I can find a way to contort my mouth in such a way that "gwitstst" gets pronounced as a single syllable.

Missing line breaks in `README.md`