🛠️ viva vocab

Viva Vocab! is a language learning game, focused on vocabulary building.

Behind the scenes it builds a model to predict a player's knowledge of words and uses it to prompt the player appropriately.

You can play it on iOS or on the web.

History

Viva Vocab was the very first app we built at Penyo Pal. Originally, it was called "Food Frenzy" (because it was just about food), but much of the original made it through to later iterations.

The first prototype was built over a weekend. Just three weeks later, a public-ready version was released when we pitched at Launch EDU in LA. We had a few thousand downloads in the first few months, but after some App Store page optimization, we started getting a few thousand downloads a week and got featured in a few regional App Stores (ahh, the early days of the app store).

That summer, we released a few other app concepts / experiments, but when the time came to pick what to jump further into, it was obviously this one.

After Penyo Pal got shuttered in early 2013, the app coasted, unchanged, eventually getting to over 1 million downloads. But since the App Store requires frequent re-uploads (say, to meet new guidelines), it got pulled and downloads stopped.

After I purchased what remained of Penyo Pal Inc in 2018, we put it back online, but to a very different mobile app ecosystem. We still get downloads and paid customers, but a far cry from the hey day of the early days. (Occasionally a school board somewhere will buy a couple hundred copies which is fun and wild).

• June 2012 - Food Frenzy
  • prototyped in a weekend, then released 3 weeks later
  • only ~ 25 food words, only Mandarin
  • competitive timer
  • already had the insight that learning image->word is seperate from word->image
  • using Moai Lua framework
• Late 2012 - Penyo Pal Episodes
  • multiple themed "episodes" which featured different word groups
  • cartoons that would play every few levels
  • casual play
  • artwork by Briana Chan
• 2016 - Viva Vocab!
  • clojurescript rewrite
  • 1000 words per language (10x more), and many more languages
  • "AI" model

Like wikisculpture, this is a project that I've worked on on-and-off for several years, and I will not let it die (despite not personally caring much about learning languages).

What keeps me coming back are:

• the positive reaction from users back in the hey day
  • and emails from fans expressing their love of it
  • (especially from ones who say they use Duolingo, but still use our app)
• the belief that we did a few legitimately valuable and novel things
  • requiring the learner to learn associations in multiple modalities (text -> image ; image -> text ; sound -> text; sound -> image)
  • avoiding native language -> new language (text -> text) prompts
  • a bayesian "AI" model that predicted a learner's knowledge of every word (and sub-challenges, like image->text) and used that to determine what words to show
• the fact that I can still speak and recognize the fruit in Mandarin, just from playing the game 13 years ago!

Plans

• more languages
• integrate a spaced repetition algorithm (that models learning decay over time)
• resurrect the android version
• play-by-voice
• other game modes (ex. memory)
• (maybe...) open source the app and the underlying data set
  • ...make it easy for people to submit new game variations