Electrical computers and digital processing systems: multicomput – Distributed data processing – Client/server
Reexamination Certificate
1999-10-19
2004-04-27
Maung, Zarni (Department: 2154)
Electrical computers and digital processing systems: multicomput
Distributed data processing
Client/server
C707S793000
Reexamination Certificate
active
06728752
ABSTRACT:
FIELD OF THE INVENTION
The invention relates to information browsing and retrieval, and more particularly to a system and method for using different document features to facilitate the identification of documents sharing one or more machine-identifiable characteristics.
BACKGROUND OF THE INVENTION
Computer users are increasingly finding navigating document collections to be difficult because of the increasing size of such collections. For example, the World Wide Web on the Internet includes millions of individual pages. Moreover, large companies' internal Intranets often include repositories filled with many thousands of documents.
It is frequently true that the documents on the Web and in Intranet repositories are not very well indexed. Consequently, finding desired information in such a large collection, unless the identity, location, or characteristics of a specific document are well known, can be much like looking for a needle in a haystack.
The World Wide Web is a loosely interlinked collection of documents (mostly text and images) located on servers distributed over the Internet. Generally speaking, each document has an address, or Uniform Resource Locator (URL), in the exemplary form “http://www.server.net/directory/file.html”. In that notation, the “http:” specifies the protocol by which the document is to be delivered, in this case the “HyperText Transport Protocol.” The “www.server.net” specifies the name of a computer, or server, on which the document resides; “directory” refers to a directory or folder on the server in which the document resides; and “file.html” specifies the name of the file.
Most documents on the Web are in HTML (HyperText Markup Language) format, which allows for formatting to be applied to the document, external content (such as images and other multimedia data types) to be introduced within the document, and “hotlinks” or “links” to other documents to be placed within the document, among other things. “Hotlinking” allows a user to navigate between documents on the Web simply by selecting an item of interest within a page. For example, a Web page about reprographic technology might have a hotlink to the Xerox corporate web site. By selecting the hotlink (often by clicking a marked word, image, or area with a pointing device, such as a mouse), the user's Web browser is instructed to follow the hotlink (usually via a URL, frequently invisible to the user, associated with the hotlink) and read a different document.
Obviously, a user cannot be expected to remember a URL for each and every document on the Internet, or even those documents in a smaller collection of preferred documents. Accordingly, navigation assistance is not only helpful, but necessary.
Accordingly, when a user desires to find information on the Internet (or other large network) that is not already represented in the user's bookmark collection, the user will frequently turn to a “search engine” to locate the information. A search engine serves as an index into the content stored on the Internet.
There are two primary categories of search engines: those that include documents and Web sites that are analyzed and used to populate a hierarchy of subject-matter categories (e.g., Yahoo), and those that “crawl” the Web or document collections to build a searchable database of terms, allowing keyword searches on page content (such as AltaVista, Excite, and Infoseek, among many others).
Also known are recommendation systems, which are capable of providing Web site recommendations based on criteria provided by a user or by comparison to a single preferred document (e.g., Firefly, Excite's “more like this” feature).
“Google” (www.google.com) is an example of a search engine that incorporates several recommendation-system-like features. It operates in a similar manner to traditional keyword-based search engines, in that a search begins by the user's entry of one or more search terms used in a pattern-matching analysis of documents on the Web. It differs from traditional keyword-based search engines (such as AltaVista), in that search results are ranked based on a metric of page “importance,” which differs from the number of occurrences of the desired search terms (and simple variations upon that theme).
Google's metric of importance is based upon two primary factors: the number of pages (elsewhere on the Web) that link to a page (i.e., “inlinks,” defining the retrieved page as an “authority”), and the number of pages that the retrieved page links to (i.e., “outlinks,” defining the retrieved page as a “hub”). A page's inlinks and outlinks are weighted, based on the Google-determined importance of the linked pages, resulting in an importance score for each retrieved page. The search results are presented in order of decreasing score, with the most important pages presented first. It should be noted that Google's page importance metric is based on the pattern of links on the Web as a whole, and is not limited (and at this time cannot be limited) to the preferences of a single user or group of users.
Another recent non-traditional search engine is IBM's CLEVER (CLient-side EigenVector Enhanced Retrieval) system. CLEVER, like Google, operates like a traditional search engine, and uses inlinks/authorities and outlinks/hubs as metrics of page importance. Again, importance (based on links/throughout the Web) is used to rank search results. Unlike Google, CLEVER uses page content (e.g., the words surrounding inlinks and outlinks) to attempt to classify a page's subject matter. Also, CLEVER does not use its own database of Web content; rather, it uses an external hub, such as an index built by another search engine, to define initial communities of documents on the Web. From hubs on the Web that frequently represent people's interests, CLEVER is able to identify communities, and from those communities, identify related or important pages.
Direct Hit is a service that cooperates with traditional search engines (such as HotBot), attempting to determine which pages returned in a batch of results are interesting or important, as perceived by users who have previously performed similar searches. Direct Hit tracks which pages in a list of search results are accessed most frequently; it is also able to track the amount of time users spend at the linked sites before returning to the search results. The most popular sites are promoted (i.e., given higher scores) for future searches.
Alexa is a system that is capable of tracking a user's actions while browsing. By doing so, Alexa maintains a database of users' browsing histories. Page importance is derived from other users' browsing histories. Accordingly, at any point (not just in the context of a search), Alexa can provide a user with information on related pages, derived from overall traffic patterns, link structures, page content, and editorial suggestions.
Knowledge Pump, a Xerox system, provides community-based recommendations by initially allowing users to identify their interests and “experts” in the areas of those interests. Knowledge Pump is then able to “push” relevant information to the users based on those preferences; this is accomplished by monitoring network traffic to create profiles of users, including their interests and “communities of practice,” thereby refining the community specifications. However, Knowledge Pump does not presently perform any enhanced search and retrieval actions like the search-engine-based systems described above.
While the foregoing systems and services blend traditional search engine and recommendation system capabilities to some degree, it should be recognized that none of them are presently adaptable to provide search-engine-like capabilities while taking into account the preferences of a smaller group than the Internet as a whole. In particular, it would be beneficial to be able to incorporate community- or cluster- based recommendations into a system that is capable of retrieving previously unknown documents from the Internet or other collectio
Chen Francine R.
Gargi Ullas
Schuetze Hinrich
Maung Zarni
Xerox Corporation
LandOfFree
System and method for information browsing using multi-modal... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for information browsing using multi-modal..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for information browsing using multi-modal... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3246509