Recommender Systems

Tell me more? The effects of mental model soundness on personalizing an intelligent agent.

by Todd Kulesza

In the proceedings of CHI 2012. Honorable mention for Best Paper award.

What does a user need to know to productively work with an intelligent agent? Intelligent agents and recommender... more What does a user need to know to productively work with an intelligent agent? Intelligent agents and recommender systems are gaining widespread use, potentially creating a need for end users to understand how these systems operate in order to fix their agent's personalized behavior. This paper explores the effects of mental model soundness on such personalization by providing structural knowledge of a music recommender system in an empirical study. Our findings show that participants were able to quickly build sound mental models of the recommender system's reasoning, and that participants who most improved their mental models during the study were significantly more likely to make the recommender operate to their satisfaction. These results suggest that by helping end users understand a system's reasoning, intelligent agents may elicit more and better feedback, thus more closely aligning their output with each user's intentions.

Towards a recommender system based on price estimation models

by Claudia Becerra

Abstract of the thesis:
The most common recommendation approaches are based on historic information. In those... more Abstract of the thesis:
The most common recommendation approaches are based on historic information. In those approaches the recommendations consist on similar products or products that have been considered by similar users. This thesis proposes a new
recommendation approach based on price product estimations that recommends those products that present the best relation value-price (“the deals”) in comparison with the other products in the same class.
Furthermore, if the price estimation model used is a linear model that calculates the contribution of each product attribute to the price, the model can be considered as a compensatory multi-attribute decision model such as a simple additive weighting model, in which initially the weights are automatically assessed.
However and given the poor prediction performance obtained using linear models as price estimation models, a new aggregated regression model was proposed called 3-Functions Membership model (3FMM) that retains the interpretability of the initial linear models at the same time that classifies the products in three price subcategories.
Following a mixture of experts scheme, the 3FMM learns how to divide the input space in 3 subspaces or categories, and for each subspace a linear price estimation model is trained. The final estimation is the average of the three estimations weighted by the membership level of the instance to each subspace.
The main motivation behind the 3FFM is to determine the inferior and superior limit in which the price oscillates and to learn in which proportion a particular instance belongs to each one of the subspaces.
The results of the experiments using the 3FMM with typical machine learning datasets, showed a decreased error margin compared with the initial model margin, when the behavior of the initial model is weak and stable. As a case study, a 1381 laptops dataset with 69 attributes was gathered from several e-commerce sites. The application of the recommendation model to this dataset made evident the usefulness of the approach in those datasets in which is possible to train a hedonic price estimation model.

User Interface Design Guidelines Arrangement in a Recommender System with Frame Ontology

by Maxim Bakaev

Lecture Notes in Computer Science, 2012, V. 7240, Database Systems for Advanced Applications, (Springer, 2012), P. 311-322.

Design guidelines, which come from the extensive body of knowledge currently formed in HCI and usability engineering... more Design guidelines, which come from the extensive body of knowledge currently formed in HCI and usability engineering domains, remain poorly integrated. Guidelines and design patterns from various sources may contradict or duplicate each other, lack links to origins and justification, as well as contextual associations to concrete problems. The paper describes how the recommender system, developed to support interface design, resolves the issues of data integration and credibility via employing frame-based ontology model and guidelines "efficiency" evaluation algorithm based on fuzzy relations. Also, experimental investigation was carried out with 24 subjects of different age groups to assess the quality of the system work. The results suggests reasonable applicability of the proposed approaches, as the success rate for the website created with the system nearly doubled the one for the control group, and guidelines efficiencies were significantly higher for relevant target user groups.

Enhancing topology adaptation in information-sharing social networks

by Giulio Cimini

Case Based Architect: CBArch

by Urjit Singh Bhatia

Citation Proximity Analysis (CPA) – A new approach for identifying related work based on Co-Citation Analysis

by Joeran Beel

Bela Gipp and Joeran Beel. Citation Proximity Analysis (CPA) - A new approach for identifying related work based on Co-Citation Analysis. In Birger Larsen and Jacqueline Leta, editors, Proceedings of the 12th International Conference on Scientometrics and Informetrics (ISSI’09), volume 2, pages 571–575, Rio de Janeiro (Brazil), July 2009. International Society for Scientometrics and Informetrics. ISSN 2175-1935. Downloaded from www.sciplore.org.

This paper presents an approach for identifying similar documents that can be used to assist scientists in finding... more This paper presents an approach for identifying similar documents that can be used to assist scientists in finding related work. The approach called Citation Proximity Analysis (CPA) is a further development of co-citation analysis, but in addition, considers the proximity of citations to each other within an article’s full-text. The underlying idea is that the closer citations are to each other, the more likely it is that they are related. In comparison to existing approaches, such as bibliographic coupling, co-citation analysis or keyword based approaches the advantages of CPA are a higher precision and the possibility to identify related sections within documents. Moreover, CPA allows a more precise automatic document classification. CPA is used as the primary approach to analyse the similarity and to classify the 1.2 million publications contained in the research paper recommender system Scienstein.org.

Recommending Mentors to Software Project Newcomers

by Igor Steinmacher

Open Source Software projects success depends on the continuous influx of newcomers and their contributions. Newcomers... more Open Source Software projects success depends on the continuous influx of newcomers and their contributions. Newcomers play an important role as they are the potential future developers, but they face difficulties and obstacles when initiating their interaction with a project, resulting in a high amount of withdrawals. This paper presents a recommendation system aiming to support newcomers finding the most appropriate project member to mentor them in a technical task. The proposed system uses temporal and social aspects of developer’s behavior, in addition to recent contextual information to recommend the most suitable mentor at the moment.

Power to the people: exploring neighbourhood formations in social recommender system

by Steven Bourke

A Simulation for Content-Based and Utility-Based Recommendation of Candidate Coalitions In Virtual Creativity Teams

by Rory Sie

Introducing serendipity in a content-based recommender system

by Michele Filannino

Can Music Be Personalized?

by Simona Profumo

Ma Interactive Media Dissertation.

The aim of this master thesis is to explore the logic of online personalization and define
how it has been... more The aim of this master thesis is to explore the logic of online personalization and define
how it has been applied for music recommendation taking in consideration the complex
nature of music. Moreover, through the analysis of several case studies of recent online
music recommender projects (AndroMedia, Lifetrak, Foafing the Music Project,
Sourcetone Project and Musicology), this paper attempts to identify the approach that
emerging trends are adopting to improve music personalization thanks to context-based
and affective recommendations.

MusicFX: an arbiter of group preferences for computer supported collaborative workouts

by Joe McCarthy

Pocket RestaurantFinder: A situated recommender system for groups

by Joe McCarthy

Designing activity-aware recommender systems for operating rooms

by Afsaneh Doryab

A recommendation system to support design patterns selection

by Alessio Malizia

Free Text In User Reviews: Their Role In Recommender Systems

by Maria Terzi

3rd Workshop on Recommender Systems and the Social Web, Held in conjunction with ACM RecSys’11 on 23rd October in Chicago, IL, USA

As short free text user-generated reviews become ubiquitous on the social web, opportunities emerge for new approaches... more As short free text user-generated reviews become ubiquitous on the social web, opportunities emerge for new approaches to recommender systems that can harness users‟ reviews in open text form. In this paper we present a first experiment towards the development of a hybrid recommender system which calculates users‟ similarity based on the content of users‟ reviews. We apply this approach to the movie domain and evaluate the performance of LSA, a state-of-the-art similarity measure, at estimating users‟ reviews similarity. Our initial investigation indicates that users‟ similarity is not well reflected in traditional score-based recommender systems which solely rely on users‟ ratings. We argue that short free text reviews can be used as a complementary and effective information source. However, we also find that LSA underperforms when measuring the similarity of short, informal user-generated reviews. For this we argue that further research is needed to develop similarity measures better suited to noisy short text.

Visualizable and Explicable Recommendations Obtained from Price Estimation Functions

by Claudia Becerra

Paper co-authored with Fabio Gonzalez and Alexander Gelbukh presented in RecSys'11 workshop on human decision making in recommender systems published on proceedings of the fifth ACM conference on Recommender systems, ACM New York, NY, USA ©2011, ISBN: 978-1-4503-0683-6 doi>10.1145/2043932.2044017

Collaborative filtering is one of the most common approaches in many current recommender systems. However, historical... more Collaborative filtering is one of the most common approaches in many current recommender systems. However, historical data and customer profileles, necessary for this approach, are not always available. Similarly, new products are constantly launched to the market lacking historical information. We propose a new method to deal with these "cold start" scenarios, designing price-estimation functions used for making recommendations based on cost-benefit analysis. Experimental results, using a data set of 836 laptop descriptions, showed that such price-estimation functions can be learned from data. Besides, they can also be used to formulate interpretable recommendations that explain to users how product features determine its price. Finally a 2D visualization of the proposed recommender system was provided.

Conception et développement de fonctionnalités innovantes liées à Facebook pour un systeme de recommandation

by Cécile Bothorel

Master 2 report

State of the art of Social Recommendation for Social TV and innovative services. State of the art of Social Recommendation for Social TV and innovative services.

Just Count the Love-Hate Squares: a Rating Network Based Method for Recommender Systems

by Kyle Teague

Recommendation Based on Opportunistic Information Sharing Between Tourists

by Michael Grossniklaus

J. of IT & Tourism 10(4): pp. 297-311 (2008)

We propose a new approach to collaborative filtering in mobile tourist information systems based on spatiotemporal... more We propose a new approach to collaborative filtering in mobile tourist information systems based on spatiotemporal proximity in social contexts. The approach is motivated by a survey of festival visitors confirming that similarities of interests extend beyond events defining specific social contexts. We show how opportunistic information sharing in mobile ad hoc networks can be used to realize decentralized collaborative filtering appropriate for mobile environments and show its equivalence to existing centralized approaches.