Thursday, May 18, 2006

GoPubMed

A colleague just pointed me to GoPubMed. This is a search engine over PubMed that organizes the returned results using GeneOntology, an ontology for molecular biology. GoPubMed takes the top abstracts from a standard keyword search, uses a soft alignment to find all the phrases in the results that have a node in GeneOntology, and then shows you the subset of GO that's induced by your query. The ontology subset lets you restrict your search in a sensible way, and get an idea of what's in documents that are fairly far down the ranked list. This looks like a nice service. It makes me wonder if we can do something similar with automatically-generated CS ontologies in Rexa.

Sunday, April 16, 2006

Rexa Launch Pictures

To celebrate Rexa's recent launch, we sponsored the weekly departmental tea at UMass this past Thursday. We demonstrated the system, and set out terminals where people could try for themselves. Food, fun, and friendly advertising were had by all. Here are some pictures:





Thursday, April 13, 2006

Rexa launch, stage 3

Today Rexa was opened to all. Anyone who visits Rexa can sign up for an account. We have also added improved topic analysis, interactive search over tags, and better format for author search. Enjoy! And feel free to spread the word.

Sunday, April 02, 2006

Rexa launch, stage 1

Rexa is a digital library and search engine covering research literature and the people who create it. Its current content emphasizes computer science research, although it also includes some statistics, mathematics, physics, bioinformatics, economics and other fields; coverage will increase in the future. Our aim is to facilitate research progress and collaboration by providing efficient browsing, search, associations and analysis and productivity tools. In this respect we are a sibling to CiteSeer, Google Scholar, and other related portals. Rexa's chief enhancement is that we use advanced statistical machine learning algorithms to perform information extraction, and then make first-class, deduplicated, cross-referenced objects not only of research papers, but also people and grants--and in the future, more objects types. This extra structure enables various new kinds of features and analysis.

The first stage of Rexa's launch happened this past week! A large handful of people were sent invitation email messages with login access. In the second stage, existing users will be able to send their own invitations (in the style of Gmail's initial launch), and in the third stage, registration access will be open to all who visit Rexa's home page. You may request an invitation by sending email to rexa-help@cs.umass.edu.