Using AltaVista LiveTopics to Search Intranets

Self-Guided Search Using LiveTopics
AltaVista's Newest Service Helps Users Pull Wheat From Chaff
Contributed by Digital Equipment Corp.

[print version of this page]

LiveTopics is a new feature on the AltaVista Search service and is breakthrough Internet search technology. It uniquely solves the problem that users often ask when faced with results indexed from the growing content on the web: "How do I make sense of the results I just got back?"

LiveTopics empowers users by enabling intelligent refinement of search queries. This it does in three ways:

Knowledge
LiveTopics brings structure and meaning to web results, results that people can use to learn about their search topic.
Relevance
LiveTopics dynamically categorizes each user's query into themes, enabling users to get to the information they are looking for faster.
Control
LiveTopics empowers users, letting them control their own search path through a complex information space.

Now in beta testing on the AltaVista Search site, users can access LiveTopics whenever an unaided search yields too many results. For result sets that come back with over 200 URL's, a new line appears on the results page called "LiveTopics". Users can then view the LiveTopics results in either HTML tables or in a Java interface that provides a visual relationship between the topics.

For example, a search for the term 'ATM' brings up references to 400,000 Web pages using that acronym. With the LiveTopics, the results appear in clear, concise topic categories, allowing a user to quickly zero in on Web pages discussing Asynchronous Transfer Mode networking, or Automatic Teller Machines.

Once AltaVista Internet Software, Inc. has completed its current beta phase of LiveTopics, it plans launch the technology worldwide across the AltaVista Search Global Affiliate Network. In addition, AltaVista will be making the technology available to its other partners and productizing the technology for Intranets.

How Does LiveTopics Work?

Think of LiveTopics as a Search Assistant that helps you find and understand information when there are too many results It reads all the results in a fraction of a second and then DYNAMICALLY categorizes the results into themes.

These themes are not predefined. In fact, they are created "on the fly" based on the results of your query. You can then use it as a learning/research tool to determine what themes are active on your topic or use it as a point-and-click advanced search tool to help you refine your query further.

Statistical analysis, not human interpretation, defines LiveTopics categories. As the Web and its content changes, the categories and search results provided by LiveTopics change automatically, making AltaVista Search results more accurate and current than competing services. Like the AltaVista Search service, LiveTopics is completely language independent: users enter search queries in their native language and the results appear in that same language.

Competing approaches

A number of companies/ search engines are trying to solve the dilemma of bringing manageability to the large amounts of unstructured data on the web. LiveTopics is unique in its approach and stands alone as a next generation search tool for the Internet.

Directories

Directories are the table-of-contents for the web. They are high-quality, low-volume, hand-crafted lists of various web sites sorted into pre-established categories. LiveTopics is serving a different mission than Directories. LiveTopics is designed to utilize AltaVista Search's entire index of the web (all pages not just sites) and dynamically categorize its contents based on each end-user's specific query.

Concept Based Searching

Basically, search technology that utilizes concept-based technology puts intelligence in a database that has automatic associations built in. It is good in that it is a smart database that tries to help users better refine their searches. However, a user has no control over it. If you are looking for something that doesn't match the database's fundamental assumptions - you will not find the web pages you are searching for.

LiveTopics is designed around on three core ideals that concept-based search technology can not deliver to end-users: 1.) Control - the end-user should have control over what they search for and how the results they get back from their searches are defined. In this way, the results that come back will better match the users intent: 2.) Relevance - no matter how many pages match a user's query, with a visual interface and an easy to use UI - LiveTopics gets people to the information they are looking for faster. The only relevant result after all, is one that the user is actually looking for: and, 3.) Knowledge - LiveTopics brings structure and meaning to web results, results that people can use to learn about their search topic No other search technology today can offer this capability.

Off-Line Topic Mapping

Off-line topic mapping technology goes out and searches several search engines at once and then fetches the pages you are interested in. Then the software does some analysis and gives you a summary sorted into pre-defined topics. This technology is good in that it brings the results of queries together into a summary form. However, it uses its own notion of categories (predefined). Also, this type of technology can only can analyze a limited number of pages as well as it is time consuming (overnight).

LiveTopics utilizes dynamic categorization technology. It actually analyzes the content on all the pages of your specific query and then categorizes those results into themes based on that specific content (Using AltaVista Search's entire web index with the results coming back usually in less than a 1/5 of a second.) The themes are dynamic based on actual content. If the web content changes, the next time you search - the resulting themes will change also.

Linguistic Processing

Linguistic Processing programs essentially reads text, analyzes the data and tries to makes sense out of it. People are using this technology to either abstract a piece of text or categorize the content into pre-established categories. The technology is still in its infancy, automatic categorization has yet to come to fruition, In addition, the technology is totally language dependent. Current implementations generally takes large amounts of CPU applied to small sets of data. Not ready to Web prime-time yet.

Automatic Categorization

This technology is a new way of mapping HTML information much like LiveTopics. The technology thus far is only designed to categorize small amounts of data and is not ready for full-scale web searching.

AltaVista plans to launch a final release version of LiveTopics across its Global Affiliate Network in Spring 1997.