|resources:||Home Mailing List Installation Source Code Members Screenshots|
Welcome to the home of MozTorch, a Mozilla machine learning extension which imbues applications with the ability to learn from past experience.
Note that this extension is intended for developers, not end-users as it provides no added functionality from the user's point of view. As such, end-users should not need to download this package unless it is required by another extension.
What's the purpose of this project? Are the author(s) obsessed with machine learning technology? Is this an excuse to play with statistical pattern recognition and get others interested? Undoubtedly. However, it also turns out that machine learning is a hugely useful field which provides promising answers to some hard problems. A few notable uses of machine learning include:
- spam filtering (e.g. SpamAssassin)
- web search (e.g. Google)
- rules of association (e.g. book suggestions on Amazon.com)
The author(s) believe there exist many more uses of machine learning techniques and that encouraging their development in the open-source community is a good way to find them. In addition... open-source development rocks, and deserves support.
An example of how MozTorch could be integrated into an existing project is the unarguably cool Monkey Do script, a GreaseMonkey snippet which automatically adds interesting pages to your del.icio.us bookmarks. The key word here is "interesting." As of this writing (8/26/2005), this script uses a pile of hand-coded heuristics (educated guesses) to classify a document as interesting. Imagine if instead the classifier could be trained by individual users to recognize their individual interests, and be corrected when it errs!
Under the hood, MozTorch interfaces with the Torch machine learning library developed by Ronan Collobert et. al. at the IDIAP Research Institute. Thanks to the IDIAP folks for providing this great open-source package!
This project has been graciously funded by Google through their Summer of Code program. The mentoring organization is Portland State University. The project mentor is Bart Massey, a friend and resource the caliber of which is rare indeed. Professor Massey coordinates all projects under the PSU mentoring organization; that's eight separate projects! Have a look at the full list of PSU projects for some very neat open-source ideas in action.