Introduction and Basic Usage
Minerva is a foreign language vocabulary drill program. Primarily designed for English speakers learning Asian languages. The initial implementation includes Simplified Chinese and Japanese vocabulary, although none of the code is specific to those two choices.
It was written as part of the 2010 Django Dash.
Users can either be registered or anonymous. Registered users can have their progress tracked over multiple sessions. Participating as an anonymous user and then registering results in progress for that session being retained as well.
The learning method is simple repetition. You are shown either an Engish meaning or a foreign word and then a choice of options in the other language. Select the correct option to match the word(s) in the question. New characters are introduced in the normal flow of questions, so you might see a character you don't know what it means yet. Pay attention to the correct answer and it will be familiar the next time it comes around.
The question selection algorithm is very rudimentary in this initial version. It's an are of almost limitless possibilities and the system design allows for experimentation there. Right now, words are graded by difficulty and easier words are presented first.
Minerva is a couple of small Django applications with only a small number of
external pre-requisites. The simplest way to get up and running quickly is
using our script to create a virtualenv environment. Assuming you already have
virtualenv and pip installed, run the
create.sh shell script in the
scripts/ directory, passing it the location of the virtual environment:
Then link in the current source code and activate the virtualenv:
cd ~/little_web_app_that_could ln -s <remember_me location> remember_me . bin/activate
The default setup uses a local SQLite database and all the data for Chinese and Japanese training is included. However, you need to create the initial database and populate the data. Continuing on from the above:
cd remember_me python manage.py syncdb # (Answer questions as prompted...) python manage.py migrate # (Wait a couple of minutes...) python manage.py runserver # (Visit http://localhost:8000/...) # (Have a rewarding beverage of choice)
For those wanting to know what's going on in the setup phase, the requirements for this set of apps are
- Django 1.2.1
- BeautifulSoup (if recreating the Japanese vocab data)
If you already have those installed, there's nothing else that needs to be done.
Obviously for something initially developed with about 20 hours of work spread over a couple of days, there are a lot of possibilities that we haven't had time to implement. These are noted here for our own future development efforts and in case people are interested.
It should be relatively straightforward to hook up with something like django-socialregistration to allow signing up via Facebook or Twitter or external registration system of choice. Not sure that socialregistration is necessarily the right choice here, but it's the type of thing we can use.
Question selection algorithm
- Things to consider
- a new user with no state, what questions are provided to him/her
- a user with no current session, would their progress be based on their progress record or current session
- Information required
- flag a word as being studied in the current session
- number of times successfully studied in the current session
- word weight
- Basic algorithm
- user submits answer
- system analyses the result and adjusts the weight
- new words may be selected if number of words is below a threshold
- Result analysis
- if answer is wrong
- push down the weight of the one they got wrong and of the word they selected
- if answer is right
push up the weight of the one they got right and increment times studied
- if number of times successfully studied is over a threshold
- Remove it from the studied item list
- New word selection
- select word with lowest attempt value
- To Consider
- retire old state