inspired by NLP Course For You 💜
WordSurfer is a playful exploration of word embedding.
WordSurfer is an terminal-based game that lets you experiment with wonderful properties of word embeddings. Built with Python's Textual framework, it currently has a functional TUI interface (with CSS styling improvements planned).
Everything in the project is built around one thing: wonderful properties of GloVe and Word2Vec (which is in core the same as based on ppmi count-based method) word embeddings (which you can read about here, here and here).
- Interactive playground for word vector arithmetic
- Quiz mode to test your intuition about word relationships
- Clean terminal interface with intuitive navigation.
-
More NLP!
Now app covers only one area of NLP but there is a lot of opportunities. For example, ai-generated localization and messages with some sentiment depending on user's achievements 😏 and more.
-
Vocabulary cleaning.
The quiz mode is full of strange combinations such as
ls + cat - dr, the result of which is really difficult to predict. This could result in bored to death community. The game needs some kind of filter. -
Design.
There is just a lot of stuff.
-
Documentation.
Currently only few functions and classes have complete docs.