Walking Randomly

There are a lot of symbols in mathematics and I mean a LOT!  Not content with the entire Greek alphabet, mathematicians have gone on to use symbols from other alphabets such as Hebrew.  Once they had run out of alphabets they went on to invent hundreds of symbols themselves – a symbol for every occasion.

So, you are writing a paper in an esoteric (or maybe not so esoteric) area of mathematics and, naturally, you are writing it in Latex.   Suddenly you think to yourself ‘What’s the LaTeX command for <insert weird and wonderful glyph here>’

Searching in vain through list after list of LaTeX symbols you get to thinking ‘If only I could just draw the symbol and have the computer tell me what the LaTeX command is‘.

Well now you can!

Detexify is a new project from Philipp Kühl (who had the initial idea) and Daniel Kirsch (who implemented it) and is essentially an exercise in machine learning.  Sometimes it works perfectly (such as in the screenshot above) but other times it struggles a bit and you end up learning the commands for symbols you never even knew existed.

Teach the system

When it is struggling though, you can help it along.  Eventually you will find the symbol you were looking for and you can click on it to tell the system ‘That squiggle I drew – this is what I meant’ thus helping to train it for future searchers.

Other times though, you cannot blame it for not finding the symbol you meant.  For example I needed about 5 tries before I could get it to recognise my ham-fisted attempt at the lowercase zeta symbol.  This says a lot more about my poor handwriting and mouse skills than it does about the quality of Texify though.

Are you rubbish with the mouse?  Use your finger on your mobile phone then!

I found drawing even simple glyphs rather difficult with the mouse and soon found myself wishing that I could do it with my finger or a stylus so I was overjoyed to learn that Robin Baumgarten has released a version of Texify for Android mobile phones.  The Android app works exactly like the web version and connects to the server in order to do the actual recognition.

Iphone users haven’t been left out though since Daniel has released an app for that himself.

This is a great project that Daniel is now developing for his diploma thesis and you and you can read more about its progress over at his blog.