wolframfriele/zamboni_python

Zamboni

This tool is a terminal based autocorrect prototype. The goal is to test out different methods of setting up autocorrection, with the goal of typing faster and learning about spelling checking / correction.

The prototype runs on curses, and needs the windows-curses library on windows.

Structure

There are some exercises you can make in the mini_projects. The folder 1_no_auto_correct, contains a blank template to have basic terminal text editor functionality, you can use this to build spelling correct into. The folder 2_single_word has autocorrect based on the edit distance to a word in a dictionary. Lastly the folder 3_context_aware contains the setup to use a markov model to improve the autocorrection.

Spelling Correct

A good wordlist to use to check words against.

Markov Chain

The markov chain can be trained on any corpus, an example of a pre tokenized dataset is the 1 billion word corpus. If you want to use a different text, it can easily be tokenized with the NLTK Tokenizer package, this functionality is not build in, but can easily be added. Look into 3_context_aware/buffer/markov_model.py in the Markov class the .split() in the _load_corpus method can be replaced with NLTK word_tokenize.