Article,
Resource recommendation in social annotation systems: A linear-weighted hybrid approach
Journal of Computer and System Sciences, 78 (4): 1160 - 1174 (2012)
DOI: 10.1016/j.jcss.2011.10.006
Abstract
Social annotation systems enable the organization of online resources with user-defined keywords. Collectively these annotations provide a rich information space in which users can discover resources, organize and share their finds, and connect to other users with similar interests. However, the size and complexity of these systems can lead to information overload and reduced utility for users. For these reasons, researchers have sought to apply the techniques of recommender systems to deliver personalized views of social annotation systems. To date, most efforts have concentrated on the problem of tag recommendation – personalized suggestions for possible annotations. Resource recommendation has not received the same systematic evaluation, in part because the task is inherently more complex. In this article, we provide a general formulation for the problem of resource recommendation in social annotation systems that captures these variants, and we evaluate two cases: basic resource recommendation and tag-specific resource recommendation. We also propose a linear-weighted hybrid framework for resource recommendation. Using six real-world datasets, we show that its integrative approach is essential for this recommendation task and provides the most adaptability given the varying data characteristics in different social annotation systems. We find that our algorithm is more effective than other more mathematically-complex techniques and has the additional advantages of flexibility and extensibility.
Tags
Users
Comments and Reviewsshow / hide
- @sdo 12 years ago (last updated 12 years ago)Topic: Item-Reco in Folksonomies through a linear combination of basic recommenders Nicely explains the basic problem formulation for different tasks (only user, user+tag, etc) and reviews several simple recommenders. Datasets: BibSonomy, CiteUlike, MovieLens, Delicious, Amazon, LastFM Methodology * Precision, Recall for different sizes of sets of recommended Items * 5-folds where each user occurs in each of the 5 folds * 1 fold for training * 4 folds for cross-validation * hold out one set and try to recommend all the resources in this held out set * hold out one set, and select a tag from that users in the held out set and try to recommend the corresponding resource * Hit-Ratio when only one resource is to be retrieved Algorithms * KNN for user based and item based CF (modeled in different vector spaces) * Vector-Space model (most similar entities to query) * Popularity * PITF (Pairwise interaction Tenosr factorization) * linear Hybrid combining all methods Results: * different on different datasets * KNN_ur on users, modelled as resource-vectors performs usually best among the simple recommenders * hybrid outperforms often the others * hybrid makes most heavily use of KNN_ur