Showing papers on "Recommender system published in 1999"

Recommender systems in e-commerce

Abstract: Recommender systems are changing from novelties used by a few E-commerce sites, to serious business tools that are re-shaping the world of E-commerce. Many of the largest commerce Web sites are already using recommender systems to help their customers find products to purchase. A recommender system learns from a customer and recommends products that she will find most valuable from among the available products. In this paper we present an explanation of how recommender systems help E-commerce sites increase sales, and analyze six sites that use recommender systems including several sites that use more than one recommender system. Based on the examples, we create a taxonomy of recommender systems, including the interfaces they present to customers, the technologies used to create the recommendations, and the inputs they need from customers. We conclude with ideas for new applications of recommender systems to E-commerce.

A Framework for Collaborative, Content-Based and Demographic Filtering

Abstract: We discuss learning a profile of user interests for recommending information sources such as Web pages or news articles. We describe the types of information available to determine whether to recommend a particular page to a particular user. This information includes the content of the page, the ratings of the user on other pages and the contents of these pages, the ratings given to that page by other users and the ratings of these other users on other pages and demographic information about users. We describe how each type of information may be used individually and then discuss an approach to combining recommendations from multiple sources. We illustrate each approach and the combined approach in the context of recommending restaurants.

Content-Based Book Recommending Using Learning for Text Categorization

Abstract: Recommender systems improve access to relevant products and information by making personalized suggestions based on previous examples of a user's likes and dislikes. Most existing recommender systems use social filtering methods that base recommendations on other users' preferences. By contrast, content-based methods use information about an item itself to make suggestions. This approach has the advantage of being able to recommended previously unrated items to users with unique interests and to provide explanations for its recommendations. We describe a content-based book recommending system that utilizes information extraction and a machine-learning algorithm for text categorization. Initial experimental results demonstrate that this approach can produce accurate recommendations.

Method and apparatus for automated selection, organization, and recommendation of items

Abstract: An automated recommendation system keeps track of the needs and preferences of the user through a user preference vector. Each field of the user preference vector represents the user's preference for a specific item attribute. Item attributes are defined by a systems programmer. The systems programmer also creates product vectors of items in the recommendation database. A user preference vector is compared against a product vector to determine if the product is suitable for recommendation. A recommended item may be purchased by the user by submitting a purchase request over a network connection. The user preference vector is constantly refined through feedback from the user about the recommended items.

Analysis of the Axiomatic Foundations of Collaborative Filtering

Abstract: The growth of Internet commerce has stimulated the use of collaborative filtering (CF) algorithms as recommender systems. Such systems leverage knowledge about the behavior of multiple users to recommend items of interest to individual users. CF methods have been harnessed to make recommendations about such items as web pages, movies, books, and toys. Researchers have proposed several variations of the technology. We take the perspective of CF as a methodology for combining preferences. The preferences predicted for the end user is some function of all of the known preferences for everyone in a database. Social Choice theorists, concerned with the properties of voting methods, have been investigating preference aggregation for decades. At the heart of this body of work is Arrow’s result demonstrating the impossibility of combining preferences in a way that satisfies several desirable and innocuous-looking properties. We show that researchers working on CF algorithms often make similar assumptions. We elucidate these assumptions and extend results from Social Choice theory to CF methods. We show that only very restrictive CF functions are consistent with desirable aggregation properties. Finally, we discuss practical implications of these results.

Clustering for collaborative filtering applications

Abstract: Collaborative ltering systems assist users to identify items of interest by providing predictions based on ratings of other users. The quality of the predictions depends strongly on the amount of available ratings and collaborative ltering algorithms perform poorly when only few ratings are available. In this paper we identify two important situations with sparse ratings: Bootstrapping a collaborative ltering system with few users and providing recommendations for new users, who rated only few items. Further, we present a novel algorithm for collaborative ltering, based on hierarchical clustering, which tries to balance robustness and accuracy of predictions, and experimentally show that it is especially e cient in dealing with the previous situations.

User profiling in personalization applications through rule discovery and validation

Abstract: Gediminas Adomavicius New York University adomavic@cs.nyu.edu In many applications, ranging from recommender systems to one-to-one marketing to Web browsing, it is important to build personalized profiles of individual users from their transactional histories. These profiles describe individual behavior of users and can be specified with sets of rules learned from user transactional histories using various data mining techniques. Since many discovered rules can be spurious, irrelevant, or trivial, one of the main problems is how to perform post-analysis of the discovered rules, i.e., how to validate customer profiles by separating “good” rules from the “bad.” This paper presents a method for validating such rules with an explicit participation of a human expert

Voting for movies: the anatomy of a recommender system

Abstract: Personal assistant agents embody a clearly beneficial application of intelligent agent technology. A particular kind of assistant agents, recommender systems (RSs), can be used to recommend items of interest to users [l]. To be successful, such systems should be able to model and reason with user preferences for items in the application domain. We are developing a movie recommender system that caters to the interests of a user. Our primary concern is to utilize a reasoning procedure that can meaningfully and systematically tradeoff between conflicting user preferences. We have adapted mechanisms from voting theory that have desirable guarantees regarding the recommendations generated from stored preferences. We provide multiple query modalities by which the user can pose unconstrained, constrained, or instance-based queries. Typically a domain has several features or dimensions. Each dimension consists of a collection of elements, and the preferences of a user are given by his/her ratings of those elements on some ordinal or cardinal scale. To obtain a recommendation rating for a given item, an RS considers the feature values of that item, obtains ratings for these values from corresponding dimensions, and then combines these ratings by some evaluation scheme.

System for providing document change information for a community of users

Abstract: A document recommendation system incorporating a document change monitoring agent. For a document recommendation system to be effective, it is desirable to enable users to be cognizant of changes that may occur to the document. The present invention addresses this issue by coupling a document change monitoring agent, which automatically detects changes in referenced documents, with a recommender system, which helps users share and evaluate information in a collaborative way. One advantage of the invention is that it brings human judgement into the relevance evaluation of the detected changes and allows the results to be shared with other people likely to be interested, in such a way that redundant work is decreased. Another advantage of the invention is that it enables Systems Administrators of a document recommendation system to be more efficient in the management of the system.

Jester 2.0 (poster abstract): evaluation of an new linear time collaborative filtering algorithm

Abstract: Jester is a WWW-based system that allows users to retrieve jokes based on their ratings of sample jokes. Our emphasis is on a new principal component analysis (PCA) and clustering-based linear time collaborative filtering algorithm for efficient and effective personalized information retrieval. Let m be the number of users in the database (currently over 12000) and n be the number of jokes rated by a user to characterize his or her preference (currently 10). We report new results comparing Jester 1.0’s O(nm) algorithm with Jester 2.0’s O(n) algorithm: the latter improves the retrieval effectiveness by more than 40% and reduces retrieval time by a factor of 12,000. To try Jester, please visit: http://shadow.ieor.berkeley.edu/humor 1. PROBLEM DEFINTION Collaborative Filtering, a.k.a. recommender systems [3], offer promising techniques for personalized information retrieval when preferences are difficult to characterize semantically [ 1,2,4]. The classic collaborative filtering problem has the Following structure given a set of objects with associated ratings, the object space is divided into two sets the predictor set and the recommendation set. A new user rates all the objects in the predictor set. Based on these ratings, objects are retrieved from the recommendation. The system should be : 1. Effective : recommended objects should receive high ratings. 2. Efficient : online recommendation process should run quickly. Konstan, Miller et. al. [2] implemented Grouplens, one of the first system for rating objects (postings) from a variety of different Usenet newsgroups including rechumor. They developed a simple but computationally intensive prediction algorithm based on weighted correlations and reported their results for a relatively small number of users. Breese et. al. [4] at Microsoft Research, classifies the collaborative filtering algorithms into two distinct classes Memory-based and Model-based: memory-based algorithms operate over the entire user database to make predictions, and model-based learn a model, which is then used for predictions. Jester 1.0 is memory-based. Jester 2.0 is modelPermission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise. to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. SIGIR ‘99 8199 Berkley, CA, USA

Augmenting recommender systems by embedding interfaces into practices

Abstract: Automated collaborative filtering systems promote the creation of a meta-layer of information, which describes users' evaluations of the quality and relevance of information items like scientific papers, books, and movies. A rich meta-layer is required, in order to elaborate statistically good predictions of the interest of the information items; the number of users' contributing to the feedback is a vital aspect for these systems to produce good prediction quality. The work presented here, first analyses the issues around recommendation collection then proposes a set of design principles aimed at improving the collection of recommendations. Finally, it presents how these principles have been implemented in one real usage setting.

Workshop on recommender systems: algorithms and evaluation

Abstract: The state of the art in recommender systems will be enhanced by the development of evaluation methodologies for recommender systems. User studies are difficult to conduct and generalize from, and issues of presentation and relevance make traditional IR evaluation measures not entirely suited to the domain. Furthermore, test collections such as DEC SRC's EachMovie data set are becoming standard tools, but the need for larger collections in different domains is great.

Improving collaborative filtering with multimedia indexing techniques to create user-adapting Web sites

Abstract: The Internet is evolving from a static collection of hypertext, to a rich assortment of dynamic services and products targeted at millions of Internet users. For most sites it is a crucial matter to keep a close tie between the users and the site.More and more Web sites build close relationships with their users by adapting to their needs and therefore providing a personal experience. One aspect of personalization is the recommendation and presentation of information and products so that users can access the site more efficiently. However, powerful filtering technology is required in order to identify relevant items for each user.In this paper we describe how collaborative filtering and content-based filtering can be combined to provide better performance for filtering information. Filtering techniques of various nature are integrated in a weighed mix to achieve more robust results and to profit from automatic multimedia indexing technologies. The combined approach is evaluated in a prototype user-adapting Web site, the Active WebMuseum.

Amalthaea and Histos: MultiAgent Systems for WWW Sites and Reputation Recommendations

Abstract: As stated in the previous chapters, recommender systems have long been a favorite application area for agent and multiagent systems researchers. The recommender systems space has traditionally been divided into systems that try to analyse the contents of documents and system that use methods other than content analysis. Content-based filtering systems are using techniques like weighed keyword vectors and SVD, while collaborative filtering and pair-wise ratings are used in the second systems category. In this chapter we will introduce two systems: Amalthaea, that falls in the former category and uses weighed keyword vectors as its representation and Histos, a pair-wise rating system that falls in the latter category.

TalkMine and the adaptive recommendation project

Abstract: TalkMine is an adaptive recommendation system which is both content-based and collaborative, and further allows the crossover of information among multiple databases searched by users. In this way, different databases learn new and adapt existing keywords to the categories recognized by its communities of users. TalkMine is based on several theories of uncertainty, as well as on biologically inspired adaptionist ideas. This system is currently being implemented for the research library of the Los Alamos National Laboratory under the Adaptive Recommendation Project. In the present work we discuss the shortcomings of current recommendation systems for distributed information systems and propose how TalkMine can greatly improve these shortcomings.

The decision-theoretic interactive video advisor

Abstract: The need to help people choose among large numbers of items and to filter through large amounts of information has led to a flood of research in construction of personal recommendation agents. One of the central issues in constructing such agents is the representation and elicitation of user preferences or interests. This topic has long been studied in Decision Theory, but surprisingly little work in the area of recommender systems has made use of formal decision-theoretic techniques. This paper describes DIVA, a decision-theoretic agent for recommending movies that contains a number of novel features. DIVA represents user preferences using pairwise comparisons among items, rather than numeric ratings. It uses a novel similarity measure based on the concept of the probability of conflict between two orderings of items. The system has a rich representation of preference, distinguishing between a user's general taste in movies and his immediate interests. It takes an incremental approach to preference elicitation in which the user can provide feedback if not satisfied with the recommendation list. We empirically evaluate the performance of the system using the EachMovie collaborative filtering database.

Intelligent information recommend system on the Internet

Abstract: In a large-scale distributed network environment like the Internet, the popularization of computers and the Internet have produced an explosion in the amount of digital information. As a result, it becomes more important and difficult to retrieve the product information adaptively to suit user preferences out of the overflowing advertisements that are changed dynamically. Though there are many types of search engines, they still have no way of adapting and filtering information for each user. To cope with these problems, we proposed a new product information recommendation system.

Interacting with recommender systems

Abstract: Many people today live in information-rich worlds, constantly facing the question: what should I do next? Which papers should I read to learn about a new area I am interested in? Which movie should I go to? Which restaurant would I like? The experience of friends and colleagues is a valuable resource for making such decisions, especially friends who are familiar with the subject area and have similar tastes.The field of recommender systems (or collaborative filtering) attempts to automate this process, e.g., by supporting people in making recommendations, finding a set of people who are likely to provide good recommendations for a given person, or deriving recommendations from implicit behavior such as browsing activity, buying patterns, and time on task.

Social and content-based information filtering for a Web graphics recommender system

Abstract: Existing social or content-based approaches to filtering-by-example are difficult to apply to image data. To realize a filtering-by-example system for image data, we propose a new approach to combine social and content-based filtering techniques. A content-based sub-system provides two types of clusters, equivalent items and virtual users, to overcome a disadvantage of social filtering, that is, a shortage of ratings. Since items similar in visual properties are not always similar in user tastes, a social sub-system controls the content-based sub-system with an evaluation function that estimates the validity of content-based clusters according to user ratings. Based on this approach, we have developed an image database, Web Graphics Navigator, that recommends graphics for Web pages according to the users' tastes. The database has been open to the public on the World-Wide Web to obtain user ratings. A preliminary observation of the user data shows promising results.

Using color and texture indexing to improve collaborative filtering of art paintings

Abstract: Information ltering is a key technology for the creation of Web sites, which are adapted to the user's needs. In this paper we identify collaborative ltering and content-based ltering as independent technologies for information ltering. We apply both technologies in our prototype user-adapting Web site, the Active WebMuseum, a recommender system for art paintings. Our new approach extends existing user pro les with content-based information gained through automatic image indexing. These extensions lead to a better performing collaborative ltering system. We validate our approach in o -line experiments.

Extending the Services and the Accessibility of Community Networks

Abstract: Community networks are community-oriented information and communication systems that are generally patterned after the public library's model of free, inclusive service and commitment to universal access. To serve the community network objectives it is therefore important to have easy and widespread information access. In this paper we present the Campiello system that proposes both enhanced information services and complementary user interfaces to better serve the community network objectives. Enhancement of the services is obtained by introducing collaborative filtering functions to support easier navigation in the information space of the community. To extend access to the community network, a paper-based interface is used, that supports exchange of information with the network, from physical locations spread in town. A large screen based interface is also used, which provides collective easy entry points to the most recent and relevant community information.

When Cookies Aren’t Enough: Tracking and Enriching Web Activity with Recer

Abstract: We describe recent work on the path model, an extension of recommender systems that focuses on the temporal order of information activity. URLs and words displayed in a web browser, as well as filenames, words and symbols used in the xemacs editor, are timestamped and logged in paths. Recommendations are made by comparing the representation of the user’s ongoing activity with the rest of their path, and with the paths of selected others. The prototype system described thus affords access to and reuse of program files, class names, variables, URLs and words consistently used in similar contexts. We treat heterogeneous data uniformly, as symbols. The Recer system relies on recurrent patterns of symbol use that emerge from activity, and that form adaptive and subjective categorisations. We describe the system’s implementation and our early experience with it, and contrast our approach with that of traditional IR. We discuss ways in which such recommender systems are set in a wider system of activity tracking, profiling, and commerce, and discuss the tradeoffs between of locality and privacy of information with its publicity and economics.

Applying Knowledge from KDD to Recommender Systems

Abstract: We investigate a new class of software for knowledge discovery in databases (KDD), called recommender systems. Recommender systems apply KDD-like techniques to the problem of making product recommendations during a live customer interaction. These systems are achieving widespread success in E-Commerce today. We extend previously studied KDD models to incorporate customer interaction so these models can be used to describe both traditional KDD and recommender systems. Recommender systems face three key challenges: producing high quality recommendations, performing many recommendations per second for millions of customers and products, and achieving high coverage in the face of data sparsity. One successful recommender system technology is collaborative filtering, which works by matching customer preferences to other customers in making recommendations. Collaborative filtering has been shown to produce high quality recommendations, but the performance degrades with the number of customers and products. New recommender system technologies are needed that can quickly produce high quality recommendations, even for very large-scale problems. For example, traditional KDD techniques might be applied in the context of our model to address these challenges. We have explored one technology called Singular Value Decomposition (SVD) to reduce the dimensionality of recommender system problems. We report an experiment where we use SVD on a recommender system database, and use the relationship between customers in the reduced factor space to generate predictions for products. We observe significant improvement in prediction quality as well as better online performance and improved coverage. Our experience suggests that SVD has the potential to meet many of the challenges of recommender systems.

Using a Painting Metaphor to Rate Large Numbers of Objects

Abstract: When retrieving information from databases or search engines, or when configuring user profiles of information filtering systems, users have to describe what objects to retrieve. While some systems require users to describe their needs using textual input, other systems simplify the users’ task by proposing a set of items, so that the user’s task is reduced to picking or rating these items (e.g. relevance feedback (Haines 93)). This task is generally much simpler than writing queries from scratch. In interfaces of this type, users have to provide information of the type “does this item represent my information interest”, “do I like this item” or “how much do I like this item”. Similar problems are encountered in utility theory, when assessing the user’s value functions (Keeney and Raiffa 76). As an example, Figure 1 shows such a selection user interface. It allows users the selection of TV channels, e.g. to configure their user profile for a TV recommender system. The interface contains about sixty toggle switches.

INVAID: an intelligent navigational AID far the world wide web

Abstract: User modelling and information filtering are two techniques that can be used for the personalisation of an information system. With user modelling, a model of a user's areas of interest can be constructed from previous interactions with the system. Two key techniques of information filtering are collaborative filtering and content based filtering. Collaborative filtering organises users with similar interests into peer groups, thus enabling the recommendation of documents considered interesting by some of the peer group to other members of that group. Content based filtering, on the other hand, compares the contents of documents with a user profile and selects those documents whose contents best match the user profile. Using a combination of these techniques, the navigational aid in the INVAID system tracks a user's navigation on the World Wide Web and subsequently recommends pages of potential interest. (3 pages)

Recommending expertise in an organizational setting

Abstract: This work explores how information systems can be augmented to assist users in finding other individuals who are likely to have specialized, expert information that they need. This paper describes a field study that considers the social and cognitive mechanisms that people use to find candidate sources of expertise. These mechanisms are the basis for a recommender system that can help users find expertise.

Social and content-based approach for visual recommendation of Web graphics

Abstract: We propose a visual browsing technique that combines content-based and social information to make use of their own advantages: specialization of filtering results and serendipitous information discovery. The system consists of a content-based map and a dynamic social filter. A visual interface based on this technique is applied to our World Wide Web graphics recommender system.

Learning User Models for Recommendation

Abstract: Recommender systems offer an unique opportunity to learn user models. Typically, a recommender system collects preference information from a large population of users and makes decisions about items that will be of greatest interest to the users. For instance, a system that recommends new movies may ask the user to list some recent movies that he or she has liked, to rate a prespecified list of movies, or to enter a few features which are important when determining what movie to see (such as the genre of the movie, favorite actors, etc.). Additionally, the system may request information about the user such as gender and age. As more information is obtained, the closer a system can approximate the user’s tastes. However, the cost of obtaining this information is very high -users generally want to minimize their interaction with a system before they are allowed to reap the benefits. It is here that we find opportunities to apply machine learning principles and derive inferences about users which could then be used in the recommendation process.

System For Recommending Items With Multiple Analyzing Components

Abstract: Relates to a recommender system using a multi-like agent, a plurality of agent like a like agent that presents the additional information for each of the recommended list of targets and each target to be recommended to the user like an independent manner; User preference information and user information of each agent holds an influential agent recommended for the user; And the User Agent selects the user preference information and the user what each using the influence information of the agent one of the plurality of like agent or more recommended agent for from, and the user preference information from the user agent, each like agents of to generate a final list of recommended target from the recommended target list that is presented from the selected recommendation agent using the additional information from the influence of the selected recommendation agent, the recommended result therefore recommended to adjust the influence of the recommended agent for the user agent retention it is characterized in that it comprises a manager. The recommendation system according to the present invention, each referral agents to create a recommendation list of targets to an independent method, presented to the recommendation manager, a variety of conventional methods like may be attempted by each like agent. In addition, like administrator recommendation from the list generated by the agent considering the influence values ​​of the additional information, user information, like agent, the final recommendation because it determines the list of targets, each like agents in the prior art a variety of like method attempts by that after that recommendations can be incorporated by the administrator, as well as recommendations, adjusting values ​​the influence of agents recommended by the reaction of the user, so it reflects the following recommendations, there is an advantage that can perform more high-quality recommendations.