Now showing 1 - 10 of 16
- PublicationExploring Customer Reviews for Music Genre Classification and Evolutionary StudiesIn this paper, we explore a large multimodal dataset of about 65k albums constructed from a combination of Amazon customer reviews, MusicBrainz metadata and AcousticBrainz audio descriptors. Review texts are further enriched with named entity disambiguation along with polarity information derived from an aspect-based sentiment analysis framework. This dataset constitutes the cornerstone of two main contributions: First, we perform experiments on music genre classification, exploring a variety of feature types, including semantic, sentimental and acoustic features. These experiments show that modeling semantic information contributes to outperforming strong bag-of-words baselines. Second, we provide a diachronic study of the criticism of music genres via a quantitative analysis of the polarity associated to musical aspects over time. Our analysis hints at a potential correlation between key cultural and geopolitical events and the language and evolving sentiments found in music reviews.
- PublicationA Distributed Asynchronous Deep Reinforcement Learning Framework for Recommender SystemsIn this paper we propose DADRL, a distributed, asynchronous reinforcement learning recommender system based on the asynchronous advantage actor-critic model (A3C), which combines ideas from A3C and federated learning (FL). The proposed algorithm keeps the user preferences or interactions on local devices and uses a combination of on-device, local recommendation models and a complementary global model. The global model is trained only by the loss gradients of the local models, rather than directly using user preferences or interactions data. We demonstrate, using well-known datasets and benchmark algorithms, how this approach can deliver performance that is comparable with the current state-of-the-art while enhancing user privacy.
- PublicationPDMFRec: A Decentralised Matrix Factorisation with Tunable User-centric PrivacyConventional approaches to matrix factorisation (MF) typically rely on a centralised collection of user data for building a MF model. This approach introduces an increased risk when it comes to user privacy. In this short paper we propose an alternative, user-centric, privacy enhanced, decentralised approach to MF. Our method pushes the computation of the recommendation model to the user’s device, and eliminates the need to exchange sensitive personal information; instead only the loss gradients of local (device-based) MF models need to be shared. Moreover, users can select the amount and type of information to be shared, for enhanced privacy. We demonstrate the effectiveness of this approach by considering different levels of user privacy in comparison with state-of-the-art alternatives.
286Scopus© Citations 20
- PublicationGenerating Personalised and Opinionated Review SummariesThis paper describes a novel approach for summarising user-generated reviews for the purpose of explaining recommendations. Wedemonstrate our approach using TripAdvisor reviews.
- PublicationA Multi-Domain Analysis of Explanation-Based Recommendation using User-Generated ReviewsThis paper extends recent work on the use of explanations in recommender systems. In particular, we show how explanations can be used to rank as well as justify recommendations, then we compare the results to more conventional recommendation approaches, in three large-scale application domains.
- PublicationAn Analysis of Recommender Algorithms for Online NewsThis paper presents the recommendation algorithms used by the Insight UCD team participating in the CLEF-NewsREEL 2014 online news recommendation challenge.
- PublicationA Live-User Study of Opinionated Explanations for Recommender SystemsThis paper describes an approach for generating rich and compellingexplanations in recommender systems, based on opinionsmined from user-generated reviews. The explanationshighlight the features of a recommended item that matter mostto the user and also relate them to other recommendation alternativesand the users past activities to provide a context.
385Scopus© Citations 33
- PublicationExplanation-based Ranking in Opinionated Recommender SystemsExplanations can help people to make better choices, but their use in recommender systems has so far been limited to the annotation of recommendations after they have been ranked and suggested to the user. In this paper we argue that explanations can also be used to rank recommendations. We describe a technique that uses the strength of an item’s explanation as a ranking signal – preferring items with compelling explanations – and demonstrate its efficacy on a real-world dataset.
- PublicationMining Product Experiences from User Generated Reviews: A Recommender Systems PerspectiveWe have employed algorithms described in to mine opinions from TripAdvisor hotel reviews; we have experimented with different parameters to learn which provided more meaningful extractions. Secondly, we have considered opinion summarization and search similar to. We have implemented a retrieval strategy that accepts natural language queries based on opinions from reviews. Additionally, we have proposed various methods of summarizing opinions based on statistical metrics. Currently, we are experimenting with feature quality metrics. Our aim is to establish a relevance score that describes the usefulness of extracted opinions. We are also running recommendation experiments using different versions the extracted opinions.
- PublicationFedFast: Going Beyond Average for Faster Training of Federated Recommender SystemsFederated learning (FL) is quickly becoming the de facto standard for the distributed training of deep recommendation models, us-ing on-device user data and reducing server costs. In a typical FLprocess, a central server tasks end-users to train a shared recommen-dation model using their local data. The local models are trained over several rounds on the users’ devices and the server combinesthem into a global model, which is sent to the devices for the pur-pose of providing recommendations. Standard FL approaches userandomly selected users for training at each round, and simply average their local models to compute the global model. The resulting federated recommendation models require significant client effortto train and many communication rounds before they converge to asatisfactory accuracy. Users are left with poor quality recommendations until the late stages of training. We present a novel technique, FedFast, to accelerate distributed learning which achieves goodaccuracy for all users very early in the training process. We achievethis by sampling from a diverse set of participating clients in each training round and applying an active aggregation method that propagates the updated model to the other clients. Consequently, with FedFast the users benefit from far lower communication costsand more accurate models that can be consumed anytime during the training process even at the very early stages. We demonstrate the efficacy of our approach across a variety of benchmark datasetsand in comparison to state-of-the-art recommendation techniques
367Scopus© Citations 41