Tuesday, April 11, 2006

The Orion Search Engine

The Orion search technology by Ori Allon seems hot recently. Someone said it was "revolutionary" and might change what people think of current search engines. I've been trying to find out what is behind Orion, but got nothing, except some reports such as the one below that describes vaguely what it is like. It seems to me that Orion offline calculates the relatedness of keywords and possible topics. One way to do that is to look at word clusters that represent various topics. Given a query, pick up the clusters that contain the query words and expand the query (or produce new sub-queries) using the word clusters to get more relevant documents. However, one word could be related to different topics represented by different word clusters. Not sure if Orion has taken into account disambiguation. Take an old example for query ambiguity, "jaguar", it would be helpful to show document clusters related to the car, the OS and the animal, then it could have related documents to each of the document clusters.

Ori Allon had been recruited by Google.

Here is the old story.

New search engine 'revolutionary'

A 26-year-old PhD student from the University of New South Wales has patented a new way of exploring the web that could revolutionise existing search engines. Developed by Ori Allon, the Orion (TM) search engine is designed to complement searches conducted on services such as Google, Yahoo or MSN Search.

Search engines find pages on which keywords occur. Sometimes these pages are important to the topic. Other times they are not.

OrionTM finds pages where the content is about a topic strongly related to the key word. It then returns a section of the page, and lists other topics related to the key word so the user can pick the most relevant.

"The results to the query are displayed immediately in the form of expanded text extracts, giving you the relevant information without having to go the website--although you still have that option if you wish," said Israeli-born Allon, who completed a Bachelor and Masters degree at Monash University in Melbourne before moving to UNSW for his PhD.

"By displaying results to other associated key words directly related to your search topic, you gain additional pertinent information that you might not have originally conceived, thus offering an expert search without having an expert's knowledge.

"Take a search such as the American Revolution as an example of how the system works. OrionTM would bring up results with extracts containing this phrase. But it would also give results for American History, George Washington, American Revolutionary War, Declaration of Independence, Boston Tea Party and more. You obtain much more valuable information from every search."

The idea of finding information without having to click through to websites came from Allon's supervisor, Eric Martin, back in March this year. "I provided the spark. But it is Ori who has developed this through his amazing creativity and sheer hard work over these past months," said Mr Martin.

Andrew Stead of New South Innovations, the technology transfer company within UNSW, says he is confident that OrionTM will fill a gap in the market noted by Microsoft founder Bill Gates.

"Bill Gates was recently quoted in Forbes magazine as saying that we need to take the search way beyond how people think of it today. We believe that OrionTM will do that."

Allon said some big companies already had shown some initial interest in implementing Orion for commercial use.

Source: University of New South Wales

Friday, April 07, 2006

A good online tutorial for vi

Found a good online tutorial for vi Vi Text Editor: Tutorial - ECN @ Purdue

Have tried to learn to use vi several times, but still can't remember how to cut, copy and paste text areas. However, I know how to do these operations in Emacs after the first use. :-) No offending to vi at all. Emacs is good for its integrated environment (you can even use Emacs as an interface for Linux/Unix). Vi is more like a handy editor that appears everywhere.