http:^^www.cs.washington.edu^research^projects^ai^590i^bs^infofilter.html
来自「This data set contains WWW-pages collect」· HTML 代码 · 共 149 行
HTML
149 行
Date: Mon, 02 Dec 1996 15:26:04 GMTServer: NCSA/1.4.2Content-type: text/html<html><head><title> Information Filtering</title></head><body><H1>Information Filtering</H1>by Guest Editors Shoshana Loch and Doug Terry <BR>From CACM December 1992, p. 49 <BR><BLOCKQUOTE><I>The promise of the information age entails making information availableto people any time, any place, and in any form. Realizing such a promisedepends on innovations in areas that impact the creation of informationservices and their communication infrastructures. However, thisrealization can easily become a mixed blessing without methods to filterand control the potentially unlimited flux of information from sources totheir receiving end-users. </I> <P></BLOCKQUOTE>Realistic deployment scenarios for information-filtering technologies havemany differentiating characteristics. For example, the type ofinformation (e.g., TV and radio programming, live news services,electronic mail), and the information transport architecture (e.g.,broadcast, narrow-cast, point-to-point) are two of the characteristicswhich strongly affect the appropriate choice of filtering technology.<P>The success of many new information services that provide end users withaccess to diverse information sources is crucially dependent on theavailability of effective filtering technology. This technology can beused by both the information sources and their end users, to route andcontrol the delivery of information. For example, in the domain ofentertainment, the individual information sources may use filters totarget material to preferred end user groups, and individual end users mayuse filters to select the material of their choice out of all availablesources. <P>The demand for information filtering technology is not new, however, andis not limited to new information services. Over a decade ago, PeterDenning's ACM President's Letter on "Electronic Junk" <CITE>(Commun.ACM, March 1982, 163-165)</CITE> focused on the implications of automaticdocument preparation systems and electronic mail, and on the quantity ofinformation being received by end users. He pointed out that "Thevisibility of personal computers, individual workstations, and local areanetworks has focused most of the attention on <I>generating</I>information--the process of producing documents and disseminating them.It is now time to focus more attention on <I>receiving</I>information--the process of controlling and filtering information thatreaches the persons who must use it." <P>In November 1991, Bellcore hosted a Workshop on High PerformanceInformation Filtering in Morristown, N.J. Organized and sponsored byBellcore in cooperation with ACM SIGOIS, the workshop was the first of itskind. The even brought together over one hundred researchers from majoruniversity and industrial research labs who share a strong interest in thecreation of large-scale personalized information delivery systems. <P>The workshop covered all aspects of this emerging area including itsrelation to the established field of information retrieval (IR), a varietyof methods for filtering, architectural concerns of high-speed filteringsystems, and a variety of existing prototype applications, as well asrequirements for future applications. <P>This special issue features five articles that represent the scope andcontent of that workshop. Each article represents a different aspect ofthe field and together they form a realistic view of the workshop. Inaddition, we present four sidebars depicting individual snapshots of anemerging filtering approach or applications. <P>Belkin and Croft ask and answer the question, "Information Filtering andInformation Retrieval: Two Sides of the Same Coin?" The authors determinethat information filtering is a well-defined process. By examining itsfoundations and comparing it to the foundations of the IR enterprise, theauthors find there is very little difference between filtering andretrieval at an abstract level. They conclude that the two enterpriseshave the same goal; namely they are both concerned with gettinginformation to people who need it. However, the authors emphasize that IRresearch has ignored some aspects of the general problem which both IR andinformation filtering address, and that these aspects are precisely thosewhich [sic] especially relevant to the specific contexts of filtering. <P>Loeb picks up where Belkin and Croft's article left off--examining some ofthe ways information-filtering models may extend IR models. Morespecifically, Loeb's article centers on "Architecting PersonalizedDelivery of Multimedia Information," providing both a mapping of thefiltering application and usage scenarios, and a specific example of anovel filtering model and its implementation. The author provides ananalysis of successful filtering applications in the context of thepersonalized multimedia music system. <P>In "Personalized Information Delivery: An Analysis of InformationFiltering Methods," Foltz and Dumais present results of an experimentaimed at determining the effectiveness of four information-filteringmethods in the domain of technical reports. The experiment was conductedover a six-month period with 34 users and over 150 new reports publishedeach month. Overall, the authors conclude that filtering methods showpromise for presenting personalized information. <P>In "Using Collaborative Filtering to Weave an Information Tapestry,"Goldberg, Nichols, Oki, and Terry describe an experimental system thatmanages an in-coming stream of electronic documents, including email,newswire stories and NetNews articles. The system implements a novelmechanism for collaborative filtering in which users annotate documentsbefore the documents are filtered. Because annotations are not availableat the time a new document arrives, the system supports continuous queriesthat examine the entire database of documents and take into account newlyintroduced annotations during the filtering process. <P>In "The Datacycle Architecture" Bowen et al., present the operatingprinciples of a fully implemented platform that supports veryhigh-performance information filtering. Key to realizing the architectureis the on-the-fly data filtering operation, which supports both expandedinformation retrieval functionality and conflict resolution for managementof changes to database contents. This article complements the others inthis section by describing an application-independent platform that embedsenough of the application semantics to adequately meet high-performancerequirements. <P>We believe that these five articles together with the sidebars capture theexcitement and quality of the work as reflected in the workshop. <P><hr>Sidebar topics are:<UL><LI> <I>Automating the Creation of Information Filters</I> by Curt Stevens (p.48)<LI> <I><!WA0><A HREF="http://www.cs.washington.edu/research/projects/ai/590i/bs/stadnyk.html">Modeling Users' Interests in Information Filters</A></I> by IreneStadnyk and Robert Krass (p.49)<LI> <I>Competitive Agents for Information Filtering</I> by Paul E.Baclace (p. 50)<LI> <I>Natural Language Understanding for Information-FilteringSystems</I> by Ashwin Ram (p. 80)</UL></body><hr><address><!WA1><A HREF="http://www.cs.washington.edu/homes/kepart/index.html"> kepart@cs.washington.edu </A></address></html>⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?