Froj

Court reporters and live captioners spend around 4 years learning expensive shorthand theory Froj codifies my theory 1) Dictionary generation 2) Theory resource 3) Can tailor to your accent!

Comment

Inspiration

Mirabai! She's an in-person captioner who made machine shorthand accessible by releasing an open source shorthand->translation engine a few years ago, and released her dictionary ("Plover theory" is the theory people have come up with to make sense of her dictionary, as it's not labeled (labeled dictionaries are very expensive since before my froject, is done by hand))

I've spent years learning and developing my machine short-hand theory. Learning resources are often hard to find, cost a lot (Magnum's dictionary is $99.00, but to get a breakdown of how to read it, the theory costs $249.00. but since many are missing explanations, there's a $29.95 monthly fee)

So I hope to open source my theory, and make sure other people don't find it as hard as I did to make their own theories.

This also means you can tailor it to your accent, I originally learnt an American theory so there was lots of frustration

What it does

During this Hackathon, I created FrojBot!, you give it a word, and it will tell you what combinations to press to reach that word. You can then ask for further information, what each combination breaks into, and what theory I used. You can also give it the shorthand combination, and it will tell you what word it maps to.

I had written many of the combinations before this Hackathon, as I've been learning steno for years, but it's only during this hackathon that I've been able to make these combinations accessible and readable (and in Discord! so it's very, very easy to access)

How we built it

Before this Hackathon I already have a short-hand theory that's taken me years to develop and learn, and write down the rules for, but during this hackathon I made it possible to generate the dictionary to then read with a Discord bot to analyse for you. It takes a pronunciation dictionary from Edinburgh's "Unisyn", and applies shorthand theory rules that look at both the spelling and the pronunciation, compounding them until you're left with a word, and each of the ways you could use shorthand to write it.

Challenges we ran into

For some words it generates hundreds and hundreds of valid entries, which means there needed to be a way to display only the entries that are worth learning It also takes hours to run and uses 100%of the CPU The pronunciation dictionary is licensed, and so cannot be included with the froject There's 120,000 words in the pronunciation dictionary, there are rules that I have not implemented, notably, loan words are very difficult, where the spelling and the pronunciation are hard for me to link.

Accomplishments that we're proud of

Lowered the cost of a fully labelled dictionary from $249.00+ Price: $99.00 to Free!!!

I had written many rules before the Hackathon, but I finally managed to implement them, and now have 108,404 unique words, but because it's shorthand, there's multiple ways to write them (823, 962 entries)

What we learned

Ungodly amount of regex, but during this Hackathon, how to make a bot to make this data accessible

What's next for Froj

I would like to add features to make it easy to customise the theory that Froj generates. At the moment it's tailored to my theory, and therefore my accent, but I would like a setup quiz that customises that.

Built With

Share this project:

Updates