A5065595661- Recommender Systems and Techniques
- Web Data Mining and Analysis
- Information Retrieval and Search Behavior
- Topic Modeling
- AI-based Problem Solving and Planning
- Expert finding and Q&A systems
- Advanced Text Analysis Techniques
- Video Analysis and Summarization
- Semantic Web and Ontologies
- Caching and Content Delivery
- Multimedia Communication and Technology
- Sentiment Analysis and Opinion Mining
- Image Retrieval and Classification Techniques
- Sports Performance and Training
- Complex Network Analysis Techniques
- Advanced Bandit Algorithms Research
- Software Engineering Research
- Data Management and Algorithms
- Explainable Artificial Intelligence (XAI)
- Stock Market Forecasting Methods
- Speech and dialogue systems
- Sports Analytics and Performance
- Data Mining Algorithms and Applications
- Artificial Intelligence in Games
- Digital Marketing and Social Media
University College Dublin
2015-2024
Dublin City University
2015-2024
Royal College of Surgeons in Ireland
2024
Insight (China)
2018-2019
Clarity Centre for Sensor Web Technologies
2009-2013
Leopardstown Park Hospital
2003-2007
École Polytechnique Fédérale de Lausanne
2007
SeaWorld Entertainment
2002-2003
Media Design School
2001
Indiana University
2001
Recommender systems have proven to be an important response the information overload problem, by providing users with more proactive and personalized services. And collaborative filtering techniques vital component of many such recommender as they facilitate generation high-quality recom-mendations leveraging preferences communities similar users. In this paper we suggest that traditional emphasis on user similarity may overstated. We argue additional factors role play in guiding...
Case-based reasoning (CBR) is an approach to problem solving that emphasizes the role of prior experience during future (i.e., new problems are solved by reusing and if necessary adapting solutions similar were in past). It has enjoyed considerable success a wide variety tasks domains. Following brief overview traditional problem-solving cycle CBR, we examine cognitive science foundations CBR its relationship analogical reasoning. We then review representative selection research past few...
Recently the world of web has become more social and real-time. Facebook Twitter are perhaps exemplars a new generation social, real-time services we believe these types service provide fertile ground for recommender systems research. In this paper focus on one key features web, namely creation relationships between users. Like recent research, view as an important recommendation problem -- given user, UT which other users might be recommended followers/followees but unlike researchers...
Recommending news stories to users, based on their preferences, has long been a favourite domain for recommender systems research. In this paper, we describe novel approach recommendation that harnesses real-time micro-blogging activity, from service such as Twitter, the basis promoting user's RSS feeds. A preliminary evaluation is carried out an implementation of technique shows promising results.
Federated learning (FL) is quickly becoming the de facto standard for distributed training of deep recommendation models, using on-device user data and reducing server costs. In a typical FL process, central tasks end-users to train shared model their local data. The models are trained over several rounds on users' devices combines them into global model, which sent purpose providing recommendations. Standard approaches use randomly selected users at each round, simply average compute model....
We describe recommender systems and especially case-based systems. define a framework in which these can be understood. The contrasts collaborative with case-based, reactive proactive, single-shot conversational, asking proposing. Within this framework, we review selection of papers from the literature, covering development over last ten years.
Mobile phones are becoming increasingly popular as a means of information access while on-the-go. users likely to be interested in locating different types content. However, the mobile space presents number key challenges, many which go beyond issues with device characteristics such screen-size and input capabilities. In particular, changing contexts location, time, activity social interactions impact on needs that arise. order offer personalized, effective services we need understand more...
Assessing user search behavior when deciding which links to follow in rank-ordered results lists.
It is likely that mobile phones will soon come to rival more traditional devices as the primary platform for information access. Consequently, it important understand emerging access behavior of Internet (MI) users especially in relation their use handsets browsing and query-based search. In this article, we describe results a recent analysis MI habits than 600,000 European users, with particular emphasis on interest We consider range factors including whether there are key differences...
Study Design Randomized controlled trial. Objectives To compare the effects of wobble board exercises with and without feedback provided through integrating movement into a computer game system, by comparing changes in postural stability motivation. Background Therapeutic exergaming systems may offer solution to poor adherence control exercise regimes improving motivation levels during performance. Methods Twenty-two healthy adults, randomly assigned an group (n = 11) 11), completed 12...
Assessing the trustworthiness of reviews is a key issue for maintainers opinion sites such as TripAdvisor. In this paper we propose distortion criterion assessing impact methods uncovering suspicious hotel in The principle that dishonest will distort overall popularity ranking collection hotels. Thus mechanism deletes significantly, when compared with removal similar set at random. This can be quantified by comparing rankings before and after deletion, using rank correlation. We present an...
User-generated reviews are a common and valuable source of product information, yet little attention has been paid as to how best present them end-users. In this paper, we describe classification-based recommender system that is designed recommend the most helpful for given product. We large-scale evaluation our approach using TripAdvisor hotel reviews, show capable suggesting superior compared number alternative recommendation benchmarks.
Collaborative filtering (CF) is a common recommendation approach that relies on user-item ratings. However, the natural sparsity of rating data can be problematic in many domains and settings, limiting ability to generate accurate predictions effective recommendations. Moreover, some CF approaches latent features are often used represent users items, which lead lack transparency explainability. User-generated, customer reviews now commonplace websites, providing with an opportunity convey...
Collaborative filtering (CF) has been successfully deployed over the years to compute predictions on items based a user's correlation with set of peers. The black-box nature most CF applications leave user wondering how system arrived at its recommendation. This note introduces PeerChooser, collaborative recommender an interactive graphical explanation interface. Users are provided visual process and opportunity manipulate their neighborhood varying levels granularity reflect aspects current...
Though computer scientists agree that conference publications enjoy greater status in science than other disciplines, there is little quantitative evidence to support this view. The importance of journal publication academic promotion makes it a highly personal issue, since focusing exclusively on papers misses many significant published by CS conferences. Here, we aim quantify the relative and papers, showing leading conferences match impact mid-ranking journals surpass bottom half Thompson...
We describe a novel, multi-task recommendation model, which jointly learns to perform rating prediction and explanation by combining matrix factorization, for prediction, adversarial sequence learning generation. The result is evaluated using real-world datasets demonstrate improved performance, compared state-of-the-art alternatives, while producing effective, personalized explanations.
In recent years, there has been an explosion of AI research on counterfactual explanations as a solution to the problem eXplainable (XAI). These seem offer technical, psychological and legal benefits over other explanation techniques. We survey 100 distinct methods reported in literature. This addresses extent which these have adequately evaluated, both psychologically computationally, quantifies shortfalls occurring. For instance, only 21% user tested. Five key deficits evaluation are...