Technical Reports

Permanent URI for this collection

https://dl.comp.nus.edu.sg/handle/1900.100/12

Browse

Browsing Technical Reports by Author "ABDESSALEM, Talel"

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    How to Find the Best Rated Items on a Likert Scale and How Many Ratings Are Enough
    (2017-06-06T02:11:13Z) LIU, Qing; BASU, Debabrota; GOEL, Shruti; ABDESSALEM, Talel; BRESSANE, Stéphane
    One of the modern pillars of collaborative filtering and recommender systems is collection and exploitation of ratings from users. Likert scale is a psychometric quantifi er of ratings popular among the electronic commerce sites. In this paper, we consider the tasks of collecting Likert scale ratings of items and of fi nding the n-k best-rated items, i.e., the n items that are most likely to be the top-k in a ranking constructed from these ratings. We devise an algorithm, Pundit, that computes the n-k best-rated items. Pundit uses the probability-generating function constructed from the Likert scale responses to avoid the combinatorial exploration of the possible outcomes and to compute the result efficiently. We empirically and comparatively evaluate with real data sets and discuss the effectiveness and efficiency of our and competing approaches. Our method is effective and competitively efficient. Selection of the best-rated items meets, in practice, the major obstacle of the scarcity of ratings. We propose an approach that learns from the available data how many ratings are enough to meet a prescribed error and recommends how many additional ratings should be proactively sought. We also empirically evaluate with real data sets the effectiveness of our method to recommend the collection of additional ratings. The results show that the approach is practical and effective.
  • No Thumbnail Available
    Item
    Top-k Queries over Uncertain Scores
    (2016-09-03) LIU, Qing; BASU, Debabrota; ABDESSALEM, Talel; BRESSAN, Stephane
    Modern recommendation systems leverage some forms of collaborative user or crowd sourced collection of information. For instance, services like TripAdvisor, Airbnb and HungyGoWhere rely on user-generated content to describe and classify hotels, vacation rentals and restaurants. By nature of such independent collection of information, the multiplicity, diversity and varying quality of the information collected result in uncertainty. Objects, such as the services o ffered by hotels, vacation rentals and restaurants, have uncertain scores for their various features. In this context, ranking of uncertain data becomes a crucial issue. Several data models for uncertain data and several semantics for probabilistic top-k queries have been proposed in the literature. We consider here a model of objects with uncertain scores given as probability distributions and the semantics proposed by the state of the art reference work of Soliman, Hyas and Ben-David. In this paper, we explore the design space of Metropolis-Hastings Markov chain Monte Carlo algorithms for answering probabilistic top-k queries over a database of objects with uncertain scores. We are able to devise several algorithms that yield better performance than the reference algorithm. We empirically and comparatively prove the eff ectiveness and effi ciency of these new algorithms.