IBM®
Skip to main content
    United States [change]    Terms of use
 
 
 
    Home    Products    Services & solutions    Support & downloads    My account    
IBM Research

Trevi

Computer Science


 Overview

Trevi is the project name for an intranet search engine being built by IBM researchers and developers in IBM's software group. At last count, over 12 people have contributed code to Trevi.

Trevi Technology Powers W3 Search

Trevi is the technology that powers IBM's intranet search. The purpose of this site is to showcase and experiment with new and innovative ideas.

How is Trevi Different from an Internet Search Engine?

  • Software and hardware for internet search engines have to scale to billions of web pages and millions of hits per day. Intranet search engines do not need to scale to this level. For example, the IBM intranet search engine runs on just 4 dual-processor PCs, unlike Google, which needs over 10,000 servers.
  • On the internet, there may be tens or even hundreds of good answers to any particular query, whereas on a corporate intranet there are often just one or two good answers. Intranet search engines have to be better at finding that needle in the haystack. Trevi uses special ranking algorithms that are tuned for intranet search to solve this problem.
  • Internet search engines treat data as merely a sequence of keywords -- they do not understand the rich semantics of the data being indexed. This problem is more tractable within the controlled environment of an intranet, which opens the door for new kinds of searches that would be impossible at the scale of the internet. For example, Trevi indexes human-resource data within IBM, allowing users to find bluepages data by entering a person's name into the search bar.
  • We believe that intranet search will become the main entry point for finding information within a company. Our ongoing work will focus on finding relevant information on intranets by understanding both the semantics of web data and the role the searcher plays within the organization. This involves better ways to crawl data sources, semantic analysis of semistructured data, and improved user interfaces for navigating search results.
 Additional Information

    About IBMPrivacyContact