Making The Cut
Surfing Buddy
Audio-Video Streaming
Making The Cut
Scientists at the Haifa
Research Laboratory have
designed a personalized Web crawler - called WebCutter - that, instead of exploring the Web exhaustively, stops in an area specified by the user and carves out a region of the Web space around that location. By doing so it gives the user a view of pages and other items in the neighborhood around a given URL.
The screen shot at right shows a WebCutter map of the region around the Lotus Domino(TM) site. Pages relevant to the user's interests - selected in advance - are highlighted in blue; the intensity of the color reflects the extent of the relevance. WebCutter has been selected to provide the site mapping technology in Lotus Notes¨/Domino solutions as part of Domino Applications.
The Haifa team developed the system, which is now undergoing beta testing, in response to a request for new products from IBM's Internet Division. "We wanted to do personalized crawling," explains Yoelle Maarek, who leads the WebCutter project. That involves searches restricted to specific concepts chosen by the user that reveal all related Web pages. "The idea is to build a map of the neighborhood around a desired Web site and cut around it."
The WebCutter permits users to carry out a series of group operations on selected sites in the map. For example, a set of documents from different nodes can be displayed and saved with complete page contents. The system also provides graphical visualization of maps in two and three dimensions.
To create a WebCutter map, the team applied a combination of information retrieval and visualization techniques. They extended Web search heuristics originally proposed by Paul De Bra of the University of Eindhoven in The Netherlands, in order to carry out "tailorable" site mapping. They developed innovative techniques for crawling the Web to permit further explorations in directions containing relevant documents and to shape Web maps according to user-specified interests. WebCutter maps thus enable users to search and browse at the same time.
Surfing Buddy
A major goal of the computer industry is finding ways to browse the Web more efficiently. IBM Research's contribution to that goal is Web Browser Intelligence, an electronic "Internet buddy" that keeps track of users' Internet activity and helps simplify Web browsing. Known as WBI (pronounced "webby"), it is based on intelligent agent technology - software that helps users and acts on their behalf. It is available for download from IBM's alphaWorks site.
"WBI is actually a confederation of agents," explains Rob Barrett of the User Systems Ergonomics Research group at the Almaden Research Center, which devised the technology. "Some monitor your Web behavior. Others modify what you see with annotations. Yet others add extra services."
WBI, which works under Windows 95®, Windows NT® or OS/2® and is browser-independent, personalizes the Web experience in several ways. They include:
- Noticing patterns in Web browsing and suggesting short cuts. Specifically, it adds a sets of links to pages frequently visited by the user.
- Automatically checking favorite Web pages for changes - such as new editions of Web magazines, and daily doses of Dilbert.
- Testing the speed of links between sites, and graphically alerting the user if those links are fast, slow or unavailable. In cases where the user has a choice of similar sites to visit, WBI can suggest which can be reached fastest.
- Remembering every site the user visits on the Web, whether bookmarked or not, thereby making it easy to return to any site. An extension in the works will automatically enable users to visit new, unseen sites they select.
- Permitting a search through previously viewed information, in the effort to locate specific sources that the user remembers only vaguely. "It gives you a new search
engine on the Web for sites you've visited," explains Barrett. The Almaden team is now developing a browsable index of key words that will work even more efficiently. "Psychologically, recognition is easier than
recall," explains Barrett.
- Letting users look back to see what routes they took previously to visit particular sites. That speeds up the search process.
- Providing a single cache and history across multiple browsers so users can switch browsers without losing information. WBI also provides connectivity through both Proxy and Socks servers, which are firewalls for restricted information.
In one broad type of use, WBI will learn patterns of Web usage well enough to predict, for example, an individual who accesses one specific site will want to access another particular site soon after. WBI will set up a "quantum link," to advise the user about the associated site. WBI technology can be applied to teaching, information access and even electronic mail. It also has potential for helping servers, as well as users, to operate in intranet and Internet situations.
Audio-Video Streaming
Users downloading files from Web sites in the conventional way face a time constraint: they can't open a file until it's entirely downloaded - a long process for complex audio and video files. One solution is streaming technology, which permits users to see the content of files that are still downloading.
IBM Research recently released its own version of streaming technology for audio and video files. Called Bamba, it is a plug-in player that streams when a file's bit rate is less than connection bit rate and downloads and plays when the stream bit rate exceeds the connection bit rate. Bamba's download time can be hundreds of times faster than current technology.
Bamba consists of a video player, a player for audio files and another player that can stream continuous audio generated in real time. It plays back video clips at rates up to 356 kilobits per second.
Bamba evolved about a year ago at the Thomas J. Watson Research Center, during development of a video phone with audio compression from the Haifa Research Laboratory and video compression from Watson's Multimedia Applications Group. "We realized we could use the same technology to stream video over phone lines in one direction," says Marc Willebeek-LeMair, manager of multimedia networking at Watson. Because a unidirectional streamer is easier to design than two-way phone links, Bamba was completed before the video phone. "The challenges were in moving extremely quickly to produce a very robust piece of code," recalls Willebeek-LeMair.
The live audio consists of three pieces: (1) the source station with live feed connection that encodes in real time and feeds; (2) the reflector, which manages independent TCP/IP connections to (3) the clients' TCP, via firewalls. The Bamba clips can also be served off standard HTTP Web servers which also go through firewalls. Unlike other audio delivery systems, the live audio does not require a proprietary server. Thus, it can be easily incorporated into existing services. Because it goes through firewalls, the technology makes an excellent foundation for talk "radio stations" that could broadcast live on the Internet.
The audio technology proved itself in action at the Masters' golf tournament in spring and the Wimbledon tennis tournament in early summer. Both audio and video appeared at the Summer Olympic Games in Atlanta. The technology also appears on IBM's alphaWorks site. That display, says Willebeek-LeMair, has provoked interest among a wide range of users.
