by George Chambers and David Sterratt
Winnow is an application for importing and organising articles from journals. It provides a means to download and store information about articles available via the PubMed database. It is possible to utilise this information to store references, view abstracts and gain access to the full text. Winnow incorporates adaptive filtering to learn what a user finds interesting and mark new articles which appear relevant.
2005-08-05: Version 0.4 has been released. This features a new maximum entropy classifier (from the MALLET toolkit) which outperforms the Naive Bayes one used previously. The size of the download is much bigger, however. We will try to make it smaller again in the next version.
Winnow requires the Java Runtime Environment version 1.4. The version for all operating systems is available as a Java jar file from the project page. Download the file to any folder. On most operating systems you should be able to double-click on the Winnow.jar icon to start it. On a UNIX-like machine you can type:
java -jar Winnow.jar
A user manual is available from the help menu.
Here are some PDF slides describing the performance of Winnow version 0.4 with the Maximum Entropy algorithm.
Here are some PDF slides describing the performance of Winnow version 0.3.
Winnow incorporates the BrowserLauncher class developed by Eric Albert.
Winnow has been developed with the financial support of the Institute for Adaptive & Neural Computation, University of Edinburgh.