Showing papers on "Recommender system published in 2000"

Explaining collaborative filtering recommendations

Abstract: Automated collaborative filtering (ACF) systems predict a person's affinity for items or information by connecting that person's recorded interests with the recorded interests of a community of people and sharing ratings between like-minded persons. However, current recommender systems are black boxes, providing no transparency into the working of the recommendation. Explanations provide that transparency, exposing the reasoning and data behind a recommendation. In this paper, we address explanation interfaces for ACF systems - how they should be implemented and why they should be implemented. To explore how, we present a model for explanations based on the user's conceptual model of the recommendation process. We then present experimental results demonstrating what components of an explanation are the most compelling. To address why, we present experimental evidence that shows that providing explanations can improve the acceptance of ACF systems. We also describe some initial explorations into measuring how explanations can improve the filtering performance of users.

Application of Dimensionality Reduction in Recommender System - A Case Study

Abstract: : We investigate the use of dimensionality reduction to improve performance for a new class of data analysis software called "recommender systems" Recommender systems apply knowledge discovery techniques to the problem of making product recommendations during a live customer interaction. These systems are achieving widespread success in E-commerce nowadays, especially with the advent of the Internet. The tremendous growth of customers and products poses three key challenges for recommender systems in the E-commerce domain. These are: 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 largescale problems. This paper presents two different experiments where we have explored one technology called Singular Value Decomposition (SVD) to reduce the dimensionality of recommender system databases. Each experiment compares the quality of a recommender system using SVD with the quality of a recommender system using collaborative filtering. The first experiment compares the effectiveness of the two recommender systems at predicting consumer preferences based on a database of explicit ratings of products. The second experiment compares the effectiveness of the two recommender systems at producing Top-N lists based on a real-life customer purchase database from an E-Commerce site. Our experience suggests that SVD has the potential to meet many of the challenges of recommender systems, under certain conditions.

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 collaborative 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 recommend 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.

Knowledge-based recommender systems

Abstract: 1. Introduction Recommender systems provide advice to users about items they might wish to purchase or examine. Recommendations made by such systems can help users navigate through large information spaces of product descriptions, news articles or other items. As on-line information and e-commerce burgeon, recommender systems are an increasingly important tool. A recent survey of recommender systems is found in The most well known type of recommender system is the collaborative-or social-filtering type. These systems aggregate data about customers' purchasing habits or preferences, and make recommendations to other users based on similarity in overall purchasing patterns. For example, in the Ringo music recommender system (Shardanand & Maes, 1995), users express their musical preferences by rating various artists and albums, and get suggestions of groups and recordings that others with similar preferences also liked. Content-based recommender systems are classifier systems derived from machine learning research. For example, the NewsDude news filtering system is a recommender system that suggests news stories the user might like to read (Billsus & Pazzani, 1999). These systems use supervised machine learning to induce a classifier that can discriminate between items likely to be of interest to the user and those likely to be uninteresting. A third type of recommender system is one that uses knowledge about users and products to pursue a knowledge-based approach to generating a recommendation, reasoning about what products meet the user's requirements. The PersonalLogic recom-mender system offers a dialog that effectively walks the user down a discrimination tree of product features.

Internet Recommendation Systems

Abstract: Several online firms, including Yahoo!, Amazon.com, and Movie Critic, recommend documents and products to consumers. Typically, the recommendations are based on content and/or collaborative filtering methods. The authors examine the merits of these methods, suggest that preference models used in marketing offer good alternatives, and describe a Bayesian preference model that allows statistical integration of five types of information useful for making recommendations: a person’s expressed preferences, preferences of other consumers, expert evaluations, item characteristics, and individual characteristics. The proposed method accounts for not only preference heterogeneity across users but also unobserved product heterogeneity by introducing the interaction of unobserved product attributes with customer characteristics. The authors describe estimation by means of Markov chain Monte Carlo methods and use the model with a large data set to recommend movies either when collaborative filtering methods ...

Expertise recommender: a flexible recommendation system and architecture

Abstract: Locating the expertise necessary to solve difficult problems is a nuanced social and collaborative problem. In organizations, some people assist others in locating expertise by making referrals. People who make referrals fill key organizational roles that have been identified by CSCW and affiliated research. Expertise locating systems are not designed to replace people who fill these key organizational roles. Instead, expertise locating systems attempt to decrease workload and support people who have no other options. Recommendation systems are collaborative software that can be applied to expertise locating. This work describes a general recommendation architecture that is grounded in a field study of expertise locating. Our expertise recommendation system details the work necessary to fit expertise recommendation to a work setting. The architecture and implementation begin to tease apart the technical aspects of providing good recommendations from social and collaborative concerns.

Collaborative filtering by personality diagnosis: a hybrid memory- and model-based approach

Abstract: The growth of Internet commerce has stimulated the use of collaborative filtering (CF) algorithms as recommender systems. Such systems leverage knowledge about the known preferences of multiple users to recommend items of interest to other users. CF methods have been harnessed to make recommendations about such items as web pages, movies, books, and toys. Researchers have proposed and evaluated many approaches for generating recommendations. We describe and evaluate a new method called personality diagnosis (PD). Given a user's preferences for some items, we compute the probability that he or she is of the same "personality type" as other users, and, in turn, the probability that he or she will like new items. PD retains some of the advantages of traditional similarity-weighting techniques in that all data is brought to bear on each prediction and new data can be added easily and incrementally. Additionally, PD has a meaningful probabilistic interpretation, which may be leveraged to justify, explain, and augment results. We report empirical results on the EachMovie database of movie ratings, and on user profile data collected from the CiteSeer digital library of Computer Science research papers. The probabilistic framework naturally supports a variety of descriptive measurements--in particular, we consider the applicability of a value of information (VOI) computation.

System and method for searching and recommending objects from a categorically organized information repository

Abstract: A search and recommendation system employs the preferences and profiles of individual users and groups within a community of users, as well as information derived from categorically organized content pointers, to augment Internet searches, re-rank search results, and provide recommendations for objects based on an initial subject-matter query. The search and recommendation system operates in the context of a content pointer manager, which stores individual users' content pointers (some of which may be published or shared for group use) on a centralized content pointer database connected to the Internet. The shared content pointer manager is implemented as a distributed program, portions of which operate on users' terminals and other portions of which operate on the centralized content pointer database. A user's content pointers are organized in accordance with a local topical categorical hierarchy. The hierarchical organization is used to define a relevance context within which returned objects are evaluated and ordered.

System and method for searching and recommending documents in a collection using shared bookmarks

Abstract: A search and recommendation system employs the preferences and profiles of individual users and groups within a community of users, as well as information derived from shared document bookmarks, to augment Internet searches, re-rank search results, and provide recommendations for documents based on a subject-matter query. The search and recommendation system operates in the context of a shared bookmark manager, which stores individual users' bookmarks (some of which may be published or shared for group use) on a centralized bookmark database connected to the Internet. The shared bookmark manager is implemented as a distributed program, portions of which operate on users' terminals and other portions of which operate on the centralized bookmark database.

Collaborative filtering with the simple Bayesian classifier

Abstract: Many collaborative filtering enabled Web sites that recommend books, CDs, movies, and so on, have become very popular on the Internet. They recommend items to a user based on the opinions of other users with similar tastes. In this paper, we discuss an approach to collaborative filtering based on the Simple Bayesian Classifier. We defme two variants of the recommendation problem for the Simple Bayesian Classifier. In our approach, we calculate the similarity between users from negative ratings and positive ratings separately. We evaluated these algorithms using databases of movie recommendations and joke recommendations. Our empirical results show that one of our proposed Bayesian approaches significandy outperforms a correlation-based collaborative filtering algorithm. The other model outperforms as well although it shows similar performance to the correlation-based approach in some parts of our experiments.

Social Choice Theory and Recommender Systems: 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.

Recommender system and method for generating implicit ratings based on user interactions with handheld devices

Abstract: A recommender system employs implicit ratings generated from monitoring user interaction with an item, such as while listening to a music track on a MP3 player or reading an electronic book. A method for generating item recommendations includes: providing an item to a device having an application for engaging a repetitive activity with the provided item, wherein the repetitive activity occurs primarily during standalone operation of the device; generating a history of user interaction with the provided item, wherein user interaction includes engaging in the repetitive activity with the provided item; transforming the history of user interactions into an implicit rating of the provided item; and using the implicit rating of the provided item to generate recommendations of other items.

Recommender system and method

Abstract: A system for providing item recommendations includes a memory, a device, responsive to a user request, for recording an item on a hardcopy medium, and a processor, for storing ratings of items and for generating recommendations for new items based on recommendation criteria. In response to the user request, the processor stores an implicit rating for the requested item in the memory, determines whether, based on the implicit rating and the recommendation criteria, to generate an item recommendation, and if the criteria for generating a recommendation is met, generates a recommendation of a new item. The recommender system may further store a representation of the recorded item in the memory. Recommendations may be based on item to item similarities, item to user similarities or user to user similarities.

OWL: A Recommender System for Organization-Wide Learning.

Abstract: We describe the use of a recommender system to enable continuous knowledge acquisition and individualized tutoring of application software across an organization. Installing such systems will result in the capture of evolving expertise and in organization-wide learning (OWL). We present the results of a year-long naturalistic inquiry into application’s usage patterns, based on logging users’ actions. We analyze the data to develop user models, individualized expert models, confidence intervals, and instructional indicators. We show how this information could be used to tutor users. Introduction Recommender Systems typically help people select products, services, and information. A novel application of recommender systems is to help individuals select ’what to learn next’ by recommending knowledge that their peers have found useful. For example, people typically utilize only a small portion of a software application’s functionality (one study shows users applying less than 10% of Microsoft Word’s commands). A recommender system can unobtrusively note which portions of an application’s functionality that the members of an organization find useful, group the organization’s members into sets of similar users, or peers (based on similar demographic factors such as job title, or similarities in command usage patterns), and produce recommendations for learning that are specific to the individual in the context of his/her organization, peers, and current activities. This paper reports research on a recommender system (Resnick & Varian, 1997) intended to promote gradual but perpetual performance improvement in the use of application software. We present our rationale, an analysis of a year’s collected data, and a vision of how users might learn from the system. We have worked with one commercial application, and believe our approach is generally applicable. The research explores the potential of a new sort of user modeling based on summaries of logged user data. This method of user modeling enables the observation of a large number of users over a long period of time, enables concurrent development of student models and individualized expert models, and applies recommender system techniques to on-the-job instruction. Earlier work is reported in Linton (1990), and Linton (1996). Kay and Thomas (1995), Thomas (1996) report on related work with a text editor in an academic environment. A recommender system to enhance the organization-wide learning of application software is a means of promoting organizational learning (Senge, 1990). By pooling and sharing expertise, recommender systems augment and assist the natural social process of people learning from each other. This approach is quite distinct from systems, such as Microsoft’s Office Assistant, which recommend new commands based on their logical equivalence to the lessefficient way a user may be performing a task. The system presented here will (1) capture evolving expertise from community of practice (Lave & Wenger 1991), (2) support less-skilled members of the community in acquiring expertise, and (3) serve as an organizational memory for the expertise it captures. In many workplaces ... mastery is in short supply and what is required is a kind of collaborative bootstrapping of expertise. (Eales & Welch, 1995, p. 100) The main goal of the approach taken in this work is to continuously improve the performance of application users by providing individualized modeling and coaching based on the automated comparison of user models to expert models. The system described here would be applicable in any situation where a number of application users perform similar tasks on networked computers 65 From: AAAI Technical Report WS-98-08. Compilation copyright © 1998, AAAI (www.aaai.org). All rights reserved. In the remainder of this section we describe the logging process and make some initial remarks about modeling and coaching software users. We then present an analysis of the data we have logged and our process of creating individual models of expertise. In the final section we describe further work and close with a summary. Each time a user issues a Word command such as Cut or Paste, the command is written to the log, together with a time stamp, and then executed. The logger, called OWL for Organization-Wide Learning, comes up when the user opens Word; it creates a separate log for each file the user edits, and when the user quits Word, it sends the logs to a server where they are periodically loaded into a database for analysis. A toolbar button labeled ’OWL is ON’ (or OFF) informs users of OWL’s tate and gives them control. Individual models of expertise We have selected the Edit commands for further analysis. A similar analysis could be performed for each type of command. The first of the three tables in Figure 1 presents data on the Edit commands for each of our 16 users. In the table, each column contains data for one user and each row contains data for one command (Edit commands that were not used have been omitted). A cell then, contains the count of the number of times the individual has used the command. The columns have been sorted so that the person using the most commands is on the left and the person using the fewest is on the right. Similarly, the rows have been sorted so that the most frequently used command is in the top row and the least frequently used command is in the bottom row. Consequently the cells with the largest values are in the upper left corner and those with the smallest values are in the lower right comer. The table has been shaded to make the contours of the numbers visible: the largest numbers have the darkest shading and the smallest numbers have no shading, each shade indicates an order of magnitude. Inspection of the first table reveals that users tend to acquire the Edit commands in a specific sequence, i.e., those that know fewer commands know a subset of the commands used by their more-knowledgeable peers. If instead, users acquired commands in an idiosyncratic order, the data would not sort as it does. And if they acquired commands in a manner that strongly reflected their job tasks or their writing tasks, there would be subgroups of users who shared common commands. Also, the more-knowledgeable users do not replace commands learned early on with more powerful commands, but instead keep adding new commands to their repertoire. Finally, the sequence of command acquisition corresponds to the commands’ frequency of use. While this last point is not necessarily a surprise, neither is it a given. There are some peaks and valleys in the data as sorted, and a fairly rough edge where commands transition from being used rarely to being used not at all. These peaks, valleys, and rough edges may represent periods of repetitive tasks or lack of data, respectively, or they may represent overdependence on some command that has a more powerful substitute or ignorance of a command or of a task (a sequence of commands) that uses the command. In other words, some of the peaks, valleys, and rough edges may represent opportunities to learn more effective use of the software. In the second table in Figure 1 the data have been smoothed. The observed value in each cell has been replaced by an expected value, the most likely value for the cell, using a method taken from statistics, based on the row, column and grand totals for the table (Howell, 1982). In the case of software use, the row effect is the overall relative utility of the command (for all users) and the column effect is the usage of related commands by the individual user. The expected value is the usage the command would have if the individual used it in a manner consistent with his/her usage of related commands and consistent with his/her peers’ usage of the command. These expected values are a new kind of expert model, one that is unique to each individual and each moment in time; the expected value in each cell reflects the individual’s use of related commands, and one’s peers’ use of the same command. The reason for differences between observed and expected values, between one’s actual and expert model, might have several explanations such as the individual’s tasks, preferences, experiences, or hardware, but we are most interested when the difference indicates the lack of knowledge or skill.

Discovery of Aggregate Usage Profiles for Web Personalization

Abstract: 1 Please direct correspondence to mobasher@cs.depaul.edu Abstract: Web usage mining, possibly used in conjunction with standard approaches to personalization such as collaborative filtering, can help address some of the shortcomings of these techniques, including reliance on subjective user ratings, lack of scalability, and poor performance in the face highdimensional and sparse data. However, the discovery of patterns from usage data by itself is not sufficient for performing the personalization tasks. The critical step is the effective derivation of good quality and useful (i.e., actionable) "aggregate usage profiles" from these patterns. In this paper we present and experimentally evaluate two techniques, based on clustering of user transactions and clustering of pageviews, in order to discover overlapping aggregate profiles that can be effectively used by recommender systems for real-time personalization. We evaluate these techniques both in terms of the quality of the individual profiles generated, as well as in the context of providing recommendations as an integrated part of a personalization engine.

Integrating Knowledge-based and Collaborative-filtering Recommender Systems

Abstract: Knowledge-based and collaborative-filtering recommender systems facilitate electronic commerce by helping users find appropriate products from large catalogs. This paper discusses the strengths and weaknesses of both techniques and introduces the possibility of a hybrid recommender system that combines the two approaches. An approach is suggested in which knowledge-based techniques are used to bootstrap the collaborative filtering engine while its data pool is small, and the collaborative filter is used as a postfilter for the knowledge-based recommender. Collaborative Filtering Collaborative filtering recommender system are a widelyaccepted technique in electronic commerce. (See [Resnick & Varian, 1997] and other articles in that special issue. A recent survey is found in [Maes, Guttman & Moukas, 1999]. See also [Goldberg et al. 1992] and [Resnick, et al. 1994].) These systems aggregate data about customers’ purchasing habits or preferences and make recommendations to other users based on similarity in overall patterns. For example, in the Ringo music recommender system (Shardanand & Maes, 1995), users who had expressed their musical preferences by rating various artists and albums could get suggestions of other groups and recordings that others with similar preferences also liked. As a collaborative filtering system collects more ratings from more users, the probability increases that someone in the system will be a good match for any given new user. This beneficial property also has its downside, however. A collaborative filtering system must be initialized with a large amount of data, because a system with a small base of ratings is unlikely to be very useful. Further, the accuracy of the system is very sensitive to the number of rated items that can be associated with a given user (Shardanand & Maes, 1995). These factors contribute to a “ramp-up” problem: until there is a large number of users whose habits are known, the system cannot be useful for most users, and until a sufficient number of rated items has been collected, the system cannot be useful for a particular user. Another problem with collaborative filtering systems might be called the “banana” problem. Bananas are a frequently-purchased item in most American grocery stores, and the odds are high that any given market basket will contain bananas. A naive recommender system working from market basket data will always recommend bananas, simply because they are highly correlated with everything. Because the system has no notion of what foods ought to go together, it cannot screen out such suggestions. These drawbacks are not significant for some large ecommerce sites, such as Amazon.com, with a very large customer base, and a large and diverse product line that lends itself to multi-item purchases. A more difficult challenge is presented for a product such as an automobile that is bought much less frequently and one at a time. For an automobile, a home loan or any other infrequentlypurchased item, the system will not be able to use marketbasket or purchase history to make recommendations. A recommender system would never be able to say “people who bought a Geo Metro also bought a Ford Escort,” because that is not how people buy cars. Knowledge-based recommender systems What a recommender system for a car or other similar product must do is get information about users’ preferences: Why are they buying a car? Is comfort or fuel economy more important? Based on such information, the system can pursue a knowledge-based approach to generating a recommendation, by reasoning about what products meet the user’s requirements. The PersonalLogic recommender system offers a dialog that effectively walks the 1 The need to maintain user-identified logs of preferences and purchases also raises privacy concerns for collaborative filtering

System and method for caching

Abstract: A system and method of caching uses quality or value attributes, provided for example, by a recommender system or by a dynamical analysis of site accesses, which are attached to cached information to prioritize items in the cache. Documents are prioritized in the cache according to the relative value of their content. Value data may be provided from a recommender system which provides a value for a document according to user recommendations (using explicit recommendations) or from statistical analysis of site visits from unique users (implicit recommendations) or a combination of the two to identify the higher value documents. The caching method may also be used to improve performance of a recommender system.

Knowledge Based Recommender Systems Using Explicit User Models

Abstract: While recommender systems are in widespread use, they still experience problems. Many recommender systems produce recommendations which the customers find unsatisfactory. Further, these systems often suffer from problems when there are not enough participants, or when new products enter the system. We perceive an opportunity for knowledge-based recommender systems to gain leverage on recommendation tasks by using explicit models of both the user of the system and the products being recommeded. This differs from previous systems which, when they use a user model, have used one that is inferred from the ratings given by that user (i.e., an implicit user model). We believe that the additional information given by the user and product models can give the system leverage in difficult recommendation tasks, and also alleviate both the "early rater" problem and the "sparse ratings" problem experienced by current recommender systems~

Hybrid Recommender Systems for Electronic Commerce

Abstract: In electronic commerce applications, prospective buyers may be interested in receiving recommendations to assist with their purchasing decisions. Previous research has described two main models for automated recommender systems - collaborative filtering and the knowledge-based approach. In this paper, we present an architecture for designing a hybrid recommender system that combines these two approaches. We then discuss how such a recommender system can switch between the two methods, depending on the current support for providing good recommendations from the behaviour of other users, required for the collaborative filtering option. We also comment on how the overall design is useful to support recommendations for a variety of product areas and present some directions for future work. Overview One of the tasks in the application area of knowledgebased electronic markets is that of providing recommendations to potential shoppers. In an environment where there is a wide choice for the prospective buyers, an automated system which serves to present a more narrow selection for the buyer would be desirable. Recommender systems are systems which provide recommendations to potential buyers. Two widely used techniques for building recommender systems to date are collaborative filtering and knowledge-based approaches. Collaborative filtering is a real-time personalization technique that leverages similarities between people to make recommendations (Greening 1998). other words, a collaborative filtering recommender system assumes that human preferences are correlated; thus, it predicts preferences and makes recommendations to one user based on the preferences of a group of users. In contrast, a knowledge-based recommender system exploits its knowledge base of the product domain to generate recommendations to a user, by reasoning about what products meet the user’s requirements. In this paper, we analyze the advantages and shortcomings of both techniques and present an architecture for a hybrid recommender system that integrates the two approaches. Such a system will inherit all the strengths from a collaborative filtering recommender system, but will be able to avoid its weaknesses. Although there have been some proposals for designing systems which make use of both the knowledge-base approach and collaborative filtering (Burke 1999), collaborative filtering is used more in a post-processing stage, so that the knowledge-based design predominates. In this paper, we outline some specifications for changing between the collaborative filtering and the knowledge-based styles of recommendation, within a single system. This design strategy will be useful for electronic commerce applications where the number of buyers and the make-up of the community of buyers dictates whether collaborative filtering will be effective or not. Background In this section, we introduce the collaborative filtering and knowledge-based approaches in building recommender systems. We discuss the strengths and weaknesses of each approach as the motivation for the design of a hybrid architecture that combines the two approaches.

Adaptation to Drifting User's Interests

Abstract: In recent years, many systems have been developed which aim at helping users to find pieces of information or other objects that are in accordance with their personal interests. In these systems, machine learning methods are often used to acquire the user interest profile. Frequently user interests drift with time. The ability to adapt fast to the current user's interests is an important feature for recommender systems. This paper presents a method for dealing with drifting interests by introducing the notion of gradual forgetting. Thus, the last observations should be more "important" for the learning algorithm than the old ones and the importance of an observation should decrease with time. The conducted experiments with a recommender system show that the gradual forgetting improves the ability to adapt to drifting user's interests. Experiments with the STAGGER problem provide additional evidences that gradual forgetting is able to improve the prediction accuracy on drifting concepts (incl. drifting user's interests).

Recommender Systems for Learning: Building User and Expert Models through Long-Term Observation of Application Use

Abstract: Information technology has recently become the medium in which much professional office work is performed. This change offers an unprecedented opportunity to observe and record exactly how that work is performed. We describe our observation and logging processes and present an overview of the results of our long-term observations of a number of users of one desktop application. We then present our method of providing individualized instruction to each user by employing a new kind of user model and a new kind of expert model. The user model is based on observing the individual's behavior in a natural environment, while the expert model is based on pooling the knowledge of numerous individuals. Individualized instructional topics are selected by comparing an individual's knowledge to the pooled knowledge of her peers.

Personalized Conversational Case-Based Recommendation

Abstract: In this paper, we describe the Adaptive Place Advisor, a user adaptive, conversational recommendation system designed to help users decide on a destination, specifically a restaurant. We view the selection of destinations as an interactive, conversational process, with the advisory system inquiring about desired item characteristics and the human responding. The user model, which contains preferences regarding items, attributes, values, value combinations, and diversification, is also acquired during the conversation. The system enhances the user's requirements with the user model and retrieves suitable items from a case-base. If the number of items found by the system is unsuitable (too high, too low) the next attribute to be constrained or relaxed is selected based on the information gain associated with the attributes. We also describe the current status of the system and future work.

TV Content Recommender System

Abstract: The plethora of content available to the consumer has become overwhelming. Increasing amounts of information are being disseminated through terrestrial broadcast, satellite, and cable leading to an information overload. Common modes of searching for TV programs currently in existence include: TV-guide, PreVue channel and rudimentary search tools available through satellite dish TV programming service. These tools are general-purpose in nature and are not specifically tailored to the individual viewer’s taste. Towards that end we advance in this paper a recommender system that searches for TV programs based on their likes/dislikes through implicit personalization techniques.

Virtual reviewers for collaborative exploration of movie reviews

Abstract: We propose a collaborative exploration system that helps users to explore recommendations from various viewpoints. Given ratings and reviews on movies from reviewers, the system provides “virtual reviewers” that represent particular viewpoints. Each virtual reviewer navigates the user by recommending and characterizing both movies and reviewers according to its viewpoint. We have developed a browsing method with virtual reviewers and visual interfaces.

Collaborative Filtering and the Generalized Vector Space Model

Abstract: Collaborative filtering is a technique for recommending documents to users based on how similar their tastes are to other users. If two users tend to agree on what they like, the system will recommend the same documents to them. The generalized vector space model of information retrieval represents a document by a vector of its similarities to all other documents. The process of collaborative filtering is nearly identical to the process of retrieval using GVSM in a matrix of user ratings. Using this observation, a model for filtering collaboratively using document content is possible.

A Case-Based Reasoning Approach to Collaborative Filtering

Abstract: Collaborative filtering systems make recommendations based on the accumulation of ratings by many users. The process has a case-based reasoning flavor: recommendations are generated by looking at the behavior of other users who are considered similar. However, the features associated with a user are semantically weak compared with those used by CBR systems. This research examines multi-dimensional or semantic ratings in which a system gets information about the reason behind a preference. Experiments show that metrics in which the semantic meaning of each rating is taken into account have markedly superior performance than simpler techniques.

Learning Subjective Functions with Large Margins

Abstract: In this paper, the authors study optimization and decision making. They propose an algorithm that is based on the theory of support vector machines. The algorithm is advantageous in that prior knowledge about the domain can be used to constrain the solution. The algorithm is demonstrated in a route recommendation system which adapts to the driver's route preferences.

SWAMI: a framework for collaborative filtering algorithm development and evaluation.

Abstract: We present a Java-based framework, SWAMI (Shared Wisdom through the Amalgamation of Many Interpretations) for building and studying collaborative filtering systems. SWAMI consists of three components: a prediction engine, an evaluation system, and a visualization component. The prediction engine provides a common interface for implementing different prediction algorithms. The evaluation system provides a standardized testing methodology and metrics for analyzing the accuracy and run-time performance of prediction algorithms. The visualization component suggests how graphical representations can inform the development and analysis of prediction algorithms. We demonstrate SWAMI on the EachMovie data set by comparing three prediction algorithms: a traditional Pearson correlation-based method, support vector machines, and a new accurate and scalable correlation-based method based on clustering techniques.

Using category-based collaborative filtering in the Active WebMuseum

Abstract: Collaborative filtering is an important technology for creating user-adapting Web sites. In general the efforts of improving filtering algorithms and using the predictions for the presentation of filtered objects are decoupled. Therefore, common measures (or metrics) for evaluating collaborative filtering (recommender) systems focus mainly on the prediction algorithm. It is hard to relate the classic measurements to actual user satisfaction because of the way the user interacts with the recommendations, determined by their representation, influences the benefits for the user. We propose an abstract access paradigm, which can be applied to the design of filtering systems, and at the same time formalizes the access to filtering results via multi-corridors (based on content-based categories). This leads to new measures which better relate to the user satisfaction. We use these measures to evaluate the use of various kinds of multi-corridors for our prototype user-adapting Web site, the Active WebMuseum.