Balabanovic, M. and Y. Shoham, 1997. Fab: Content-based, collaborative recommendation. Communications of the ACM, 40: 66-72.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....into the use of NLP techniques to aid recommendation systems. There may also be advantages of combining these similarity metrics with collaborative filtering approaches. We believe that using such content enhanced approaches will not only improve recommendation quality, as suggested in [1,6], but also will lead to increased functionality; namely, the ability to ask the system for suggestions in a novel way, e.g. Suggest a movie I will like that is similar to X . Thus, while this is only a first step toward exploiting NLP in recommendations, we believe it is a promising and necessary ....

Balabanovic, M., and Shoham, Y. Fab: Content based, collaborative recommendation. Communications of the ACM, 40(3), (1997), 66-72.

....managed by site managers: techniques of collaborative filtering can be introduced where user profiles and evaluations are stored and used to automatically build a list of possibly attractive links specifically tailored for a particular user. Many recommendation systems, such as Tapestry [4] or Fab [2], require users to express their evaluation of the visited item, while others can gather implicit information. For example, the GroupLens [7] USENET news recommendation system uses reading times as a user interest measure. Other system, such as PHOAKS [13] use data mining techniques to extract ....

M. Balabanovic and Y. Shoham. Fab: Content-based, collaborative recommendation. Communications of the ACM, 40(3):66--72, Mar. 1997.

....recommender systems. Goldberg et al. 17] define collaborative filtering as collaboration in which people help one another perform filtering by recording their reactions to Web documents they read. Examples of collaborative filtering and recommender systems include Amazon.com, GroupLens [26] Fab [3], Ringo [38] and Do I Care [39] When a user performs a search, these systems will recommend a set of documents or items that may be of interest based on that user s profile and other users interests and past actions. For example, Amazon. com uses collaborative filtering to recommend books to ....

....Agent keeps a list of monitoring tasks and is responsible for carrying out these tasks based on users schedules. Our architecture differs from traditional information retrieval systems or recommender systems in that collaboration is based on users searches and analysis, not Web pages rated [3], news articles viewed [26] or items purchased (e.g. Amazon.com) The functionalities of each type of agent are discussed in detail in the following sections. 3.1. User agent The User Agent is developed based on Competitive Intelligence (CI) Spider, a prototype system developed in our previous ....

M. Balabanovic, Y. Shoham, Fab: content-based, collaborative recommendation, Communications of the ACM 40 (3) (1997) 66 -- 72.

....further assistance know to whom they should turn for help. Programmers who are willing to contribute a little bit can activate another added Emacs command to rate the example based on its usefulness in illustrating the use of the component. The rating, which is interpreted as peer recommendation [1], goes to a database and is used by Illustrator to determine which example should be presented when several examples that use the same component exist. If ratings are available, Illustrator chooses the example with the highest average rating; otherwise, it chooses the most simple example program. ....

Balabanovic, M., and Shoham, Y. Fab: Content-Based, Collaborative Recommendation. Commun. ACM, 1997. 40(3):66-72.

....In the area of literature services personalization is mostly used for notification on new documents. Meta systems have neither access to content nor uniform information about classification of documents, so personalization techniques like content based and collaborative filtering [4] [5], 6] are not suitable. Response time: For scalability and autonomy reasons a meta search system accesses the underlying services by using their public Web interfaces. But response times of Web requests can easily take some seconds. Due to information splitting strategies of literature services ....

Balabanovic, M., Shoham, Y.: Fab: Content-based, Collaborative Recommendation. In: CACM, 40(3), p. 66-72, March 1997.

.... to any useful feature extraction with current technology (such as movies, music, restaurants) Even for text documents the representations capture only certain aspects of the content, and there are many others that would influence a user s experience, e.g. in how far it matches the user s taste [BS97] Collaborative filtering (CF) is an approach to overcome this limitation. The basic concept of CF [GNOT92] is to automate social processes such as word of mouth. In everyday life, people rely on the recommendations from other people either by word of mouth, recommendation letters, movie and ....

....described in 2.1.3, so that this method cannot correctly represent multiple interests. Mock partially solves this problem by adding the online lexical reference system WordNet that helps to identify concepts in the examined messages. The Web page recommendation service Fab by Marko Balabanovic [BS97] uses user profiles similar to those used by INFOS. Profiles, here called selection agents, are simple keyword vectors and are generated by adding the weighted keyword vectors of the training examples. Consequently, Fab suffers from similar problems as the INFOS system when trying to learn ....

[Article contains additional citation context not shown here]

M. Balabanovic and Y. Shoham, Fab: content-based, collaborative recommendation. Communications of the ACM, 40(3):66-72, March 1997.

....form the clusters and the quality of results. For example, when clustering 142 days of data, K = 200 took twice as long to form the clusters as K = 100, and K = 400 took 4 times as long. 5. RELATED WORK There are many systems which involve users ranking or judging Web sites as they visit them; [9, 2] are examples. We feel that systems which require the users to explicitly comment on the Web sites place too burden on the user. Therefore, while the data can be considerably cleaner than our search sessions, it is necessarily of limited coverage. The database in [9] for example, contains very ....

M. Balabanovic and Y. Shoham. Fab: content-based, collaborative recommendation. Communications of the ACM, 40(3):66--72, Mar. 1997.

....Scatter Gather [15] and Dynamic Taxonomies [57] projects rely on defining a set of operations under which transformations made on an information space are closed. These projects concentrate on retrieval and navigation, respectively. While there has been considerable research in web personalization [1, 2, 3, 26, 35, 40, 60], many of these algorithms systems (or in some cases their results) are usefully viewed as modeling choices to be made in a PIPE implementation. For instance, the graph theoretic recommendation algorithm described in [2] can be modeled as a function in PIPE, so that the results of the function are ....

M. Balabanovic and Y. Shoham. Fab: Content-Based, Collaborative Recommendation. Communications of the ACM, Vol. 40(3):pages 66--72, 1997.

....systems that use CBF method is precision of estimates and also recommendations. There is a hypothesis that combination of CF and CBF method can alleviate mentioned drawbacks of initial phase and make better recommendations. There were presented several approaches for combining CF and CBF methods ([3], 8] 5] 1] We propose a method for CF and CBF combination (Appendix C) where CBF estimates are used to ll up some missing ratings for CF method (Figure 3) To do this, both pro les those based on item s content (CBF) and those based on ratings (CF) are needed. Our method tries to ....

M. Balabanovic, and Y. Shoham. Fab: Content-based, Collaborative Recommendation. Communications of the ACM, 40(3):29-38, 1997.

.... approach or a strict collaborative filtering approach, recent research has shown that predictive accuracy can be greatly improved by using the methods in the inductive classification framework and learning on explicit social features (e.g. Jane liked Titanic ) as well as content based features [11,12,54]. Finally, the ReferralWeb [64] offers an interesting twist on Internet search engines rather than link people to authoritative Web pages, ReferralWeb aims to direct people (or their email questions) to humans who are experts on a given topic. Naturally, this casts a new spin on user modeling ....

M. Balabanovic, Y. Shoham, Fab: Content-based, collaborative recommendation, 1997.

....details on this case study can be obtained from [PLR00] 5 Discussion 5.1 Related Research As a systematic methodology for personalization, PIPE is a unique research project. Most research on personalization emphasizes the nature of information being modeled [RV97, THA 97] content based [BS97] versus collaborative [AT01, AWWY99, KMM 97, RP97] the level at which the personalized information is targeted (is it by user [MPR00] by topic [PE00] or for everybody [HH94, WM99] or the specific algorithms that are involved in making recommendations. In contrast, PIPE models ....

M. Balabanovic and Y. Shoham. Fab: Content-Based, Collaborative Recommendation. Communications of the ACM, Vol. 40(3):pages 66--72, 1997.

....to Usenet Newsgroup article routing. After learning from user judgments of previously read news articles, the system scours through a wide variety of newsgroups collecting other relevant articles for the user. Similarly, Balabanovic has developed a system, known initially as LIRA and later as Fab [9, 8], that works in much the same fashion for recommending potentially interesting Web pages to users. Fab s underlying technology differs from NewsWeeder in that it makes use of a collection of distributed agents to learn multiple topics of interest for each user of the system. Also working in the ....

Balabanovic, M., and Shoham, Y. Fab: Content-based, collaborative recommendation. Communications of the ACM 40, 3 (1997).

....or movies to site visitors. These recommendation mechanisms are usually based on collaborative filtering [Resnick et al. 1994; Shardanand and Maes, 1995; Hill et al. 1995 ] content based filtering [Lang, 1995; Pazzani et al. 1996; Mooney et al. 1998 ] or a combination of these two methods [Balabanovic and Shoham, 1997; Pazzani, 1999] However, in many applications, such as recommending vacation packages or restaurants to customers, it is not sufficient to recommend items to users or users to items. For example, customer preferences for specific vacation packages can depend significantly on the time of the ....

M. Balabanovic and Y. Shoham. Fab: Content-based, collaborative recommendation. Communications of the ACM, 40(3):66--72, 1997.

No context found.

Balabanovic, M. and Y. Shoham, 1997. Fab: Content-based, collaborative recommendation. Communications of the ACM, 40: 66-72.

No context found.

M. Balabanovic and Y. Shoham. Fab: content-based, collaborative recommendation. Communications of the ACM, 40(3):66--72, 1997.

No context found.

Balabanovic, M., & Shoham, Y. (1997). Fab: Content-based, collaborative recommendation. Communications of the ACM, 40(3), 66 --72.

No context found.

Balabanovic, M., Shohan, Y. (1997): Fab: Content-Based, Collaborative Recommendation. Communications of the ACM 40:3, 66--72. http://fab.stanford.edu/

No context found.

M. Balabanovic and Y. Shoham. Fab: Content-Based, Collaborative Recommendation. Communications of the ACM, 40(3):66--72, 1997.

No context found.

Balabanovic, M. and Shoham Y. Fab: Content-based, collaborative recommendation. Communications of the ACM 40, 3 (March 1997), 66-72.

No context found.

M. Balbanovic and Y. Shoham. Fab -- Content-Based, Collaborative Recommendation. Communications of the ACM, 40(3): 66-72, March 1997.

No context found.

M. Balabanovic and Y. Shoham. Fab: Content-based, collaborative recommendation. Communications of the ACM, 40(3):66--72, 1997.

No context found.

M. Balabanovic and Y. Shoham. Fab: Content-based, collaborative recommendation. Communications of the ACM, 40(3):66--72, 1997.

No context found.

Balabanovic, M., and Shoham, Y. 1997. Fab: Contentbased, collaborative recommendation. Communications of the ACM, 40(3), 66-72.

No context found.

Balabanovic, M. and Shoham, Y.: Fab: Content-based, collaborative recommendation. Communications of the ACM, 40(3):66-72, (1997).

No context found.

Balabanovic, M., Shoham, Y.: Fab - content-based, collaborative recommendation. Communications of the ACM 40 (1997) 66--72

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC