– The system can generally detect the ‘main text’ of ‘meaningful web pages’, such as newspaper articles, blog entries, magazine articles, etc., most of the time. It uses a very simple heuristic and implements this as a jquery selection. Not as good as natural intelligence but good enough for development purposes.
– By using an English word frequency list, the system detects if the ‘main text’ has a high probability of being in English. (By incorporating frequency lists for other languages such as German, French, etc. we will scale this functionality to other languages as well). It does this by comparing the current page’s word frequency with the top 20 words taken from the British National Corpus statistics.
– Then it proceeds to apply a cloze test strategy to the detected ‘main text’. The current strategy is STRATEGY_RANDOM which is basically something like ‘remove every Nth word from the text and replace it with a pull-down list and a constant distractor’. (Just as a proof of concept and debugging purposes, not as a linguistically and pedagogically meaningful test strategy at all).
Currently the plug-in is activated by a very simple way, by clicking on the Run ClozeFox! button on the bottom right corner of the browser status bar and some of the results look like these:
There is still a long way to go to reach something like that:
The source code repository is at http://github.com/emres/clozefox.
Now the development is moving in 2 parallel tracks, one being the linguistic part and the other is user interface / social collaboration part. We are about to implement meaningful and useful language test strategies and in the meantime we are exploring the JetPack platforms features for creating a UI and collaboration functionality.