Skip to content

Latest commit

 

History

History
72 lines (60 loc) · 2.66 KB

README.md

File metadata and controls

72 lines (60 loc) · 2.66 KB
Raw

Memorize App

This is an app designed to help with memorization through mnemonics. The app is not finished right now but the core functionality still works. The idea is to find words that sound like a word that you are trying to memorize. For example, instead of memorizing that huevos <means> eggs you would memorize it using a mnemoic huevos <soundslike> oasis <image> eggs with "eggs in an oasis" being the imagery. This techinque isn't for everyone but it helps me personally so I think at least some others would enjoy using the app.

How to Run

The code base comes with a management script. The only requirements are an installation of Python 3, Click and docker-compose. To generate the embeddings and feature table required run the following command: python3 manage.py dev generate

This will take a few minutes and will create the files nbow.p and embeddings.p. To start the program after this is done run: python3 manage.py dev up

Right now only the API is functional, which is a separate container from the frontend and backend. To use the api you can run a curl command to interact with it. There is currently only one endpoint:

alon@alon-XPS:~/memorize$ curl -d '{"lang" : "es", "word":"huevos"}' -H "Content-Type: application/json" -X POST http://localhost/api/mnemonic/
[
  [
    "oasis", 
    1.591621994972229
  ], 
  [
    "waves", 
    1.600000023841858
  ], 
  [
    "whatever", 
    1.600000023841858
  ], 
  [
    "paso", 
    1.632455587387085
  ], 
  [
    "waste", 
    1.6485282182693481
  ], 
  [
    "weights", 
    1.6485282182693481
  ], 
  [
    "sofa", 
    1.7638263702392578
  ], 
  [
    "survey", 
    1.8142136335372925
  ], 
  [
    "slowly", 
    1.880983829498291
  ], 
  [
    "solely", 
    1.880983829498291
  ]
]

The endpoint prints a list of similar sounding words. It should work with any language in http://espeak.sourceforge.net/languages.html list by modify the lang parameter in the API call. Some fun words to try in spanish:

  • Barato
  • Camiseta
  • Jugar

How it Works

Glad you asked! The app generates a phonetic representation of the given word using the IPA (International Phonetic Alphabet). Each phoneme in the IPA gets converted to a vector of 22 features (there are about 100 different phonemes). Then we look for words that have a similar array of vectors. This is done using Earth Mover's Distance - a way of finding similar arrays given a large list. We find similar words from a list of the top 10,000 most common english words.

The best way to see more about how the code works is to check out the mnemonic/generate.py file. This is what generates the feature arrarys from the list of 10,000 english words in the repo.