Of academic interest

MIT teaches an AI course on Commonsense Reasoning that looks fun. The information is apropos since I've thought about writing a Firefox extension that would scan through linked RSS feeds on web pages/blogs and find other blogs/articles with similar content. In other words, the extension would make Firefox into a "smart" browser that could suggest other linked articles or feeds that I would be interested in. There's way too much content out there for me than I have time to scan through.

I think the two main components to this would be a word/phrase classifier and a commonsense database. The classifier would pick up the most common occurrences of words or phrases on the web pages I frequent and be able to suggest relevant content. The commonsense database would be necessary to distinguish the context of pages, e.g. being able to tell the difference between "Java" the programming language and "java" as coffee.

Looking at MIT's page, there are apparently three main, open-source commonsense databases available on the Web: ConceptNet, ThoughtTreasure and OpenCyc. The only one I've really played with to any extent is ConceptNet because it has a slick database browser written in Python and Tk. Unfortunately, I don't have a good impression of its capability to understand most sentences, especially when there's technical content involved. It's also slow.

I'm not sure if this programming project will make it off the ground. But unlike most of my other ideas, I don't see anyone else who's working on something similar, and I have a feeling even a half-smart browser would be useful.