We present the Height Optimized Trie (HOT), a fast and space-efficient in-memory index structure. The core algorithmic idea of HOT is to dynamically vary number bits considered at each node, which enables consistently high fanout thereby good cache efficiency. layout node carefully engineered for compactness search using SIMD instructions. Our experimental results, use wide variety workloads data sets, show that outperforms other state-of-the-art structures string keys both in terms...

The extraction of information from online social networks has become popular in both industry and academia as these data sources allow for innovative applications. However, the area music recommender systems retrieval, respective is hardly exploited. In this paper, we present #nowplaying dataset, which leverages media creation a diverse constantly updated describes listening behavior users. For rely on Twitter, frequently facilitated posting user currently to. From such tweets, extract track...

New distribution channels like music streaming platforms paved way for making more and diverse available to users. Thus, recommender systems got in the focus of research academia as well industry. Collaborative filtering-based have been proven useful, but there is space left improvements by adapting this general approach better fit recommendations problem. In work, we incorporate context-based information about consumption into recommendation process. This extracted from playlist names,...

Music recommender systems have become an integral part of music streaming services such as Spotify and Last.fm to assist users navigating the extensive collections offered by them. However, while listeners interested in mainstream are traditionally served well systems, beyond (i.e., non-popular music) rarely receive relevant recommendations. In this paper, we study characteristics beyond-mainstream analyze what extent these impact quality recommendations provided. Therefore, create a novel...

Abstract Music is ubiquitous in our everyday lives, and lyrics play an integral role when we listen to music. The complex relationships between lyrical content, its temporal evolution over the last decades, genre-specific variations, however, are yet be fully understood. In this work, investigate dynamics of English Western, popular music five decades genres, using a wide set descriptors, including complexity, structure, emotion, popularity. We find that pop have become simpler easier...

Online social networks like Facebook or Twitter have become powerful information diffusion platforms as they attracted hundreds of millions users. The possibility reaching users within these not only standard users, but also cyber-criminals who abuse the by spreading spam. This is accomplished either creating fake accounts, bots, cyborgs hacking and compromising accounts. Compromised accounts are subsequently used to spread spam in name their legitimate owner. work sets out investigate how...

Recommender systems, like other tools that make use of machine learning, are known to create or increase certain biases. Earlier work has already unveiled different performance recommender systems for user groups, depending on gender, age, country, and consumption behavior. In this work, we study bias in terms another aspect, i.e., users' personality. We investigate which extent state-of-the-art recommendation algorithms yield accuracy scores the personality traits. focus music domain a...

In session-based recommender systems, predictions are based on the user's preceding behavior in session. State-of-the-art sequential recommendation algorithms either use graph neural networks to model sessions a or leverage similarity of by exploiting item features. this paper, we combine these two approaches and propose novel method, Graph Convolutional Network Extension (GCNext), which incorporates features directly into representation via convolutional networks. GCNext creates...

Music streaming platforms enable people to access millions of tracks using computers and mobile devices. The latter allow users consume different music during activities. Both, the sheer amount makes organization an interesting topic for multimedia researchers. Assisting organize their make they like easily available in right moment, contributes increased usability platforms. To get a deeper understanding how nowadays, we analyze user-created playlists crawled from platform Spotify. Using...

When we appreciate a piece of music, it is most naturally because its content, including rhythmic, tonal, and timbral elements as well lyrics semantics. This suggests that the human affinity for music inherently content-driven. kind information is, however, still frequently neglected by mainstream recommendation models based on collaborative filtering rely solely user-item interactions to recommend items users. A major reason this neglect lack standardized datasets provide both content...

In this paper, we focus on recommendation settings with multiple stakeholders possibly varying goals and interests, argue that a single evaluation method or measure is not able to evaluate all relevant aspects in such complex setting. We reason employing multi-method evaluation, where methods measures are combined integrated, allows for getting richer picture prevents blind spots the outcome.

Mood and emotion play an important role when it comes to choosing musical tracks listen to. In the field of music information retrieval recommendation, is considered contextual that hard capture, albeit highly influential. this study, we analyze connection between users` emotional states their choices. Particularly, perform a large-scale study based on two data sets containing 560,000 90,000 #nowplaying tweets, respectively. We extract affective from hashtags contained in these tweets by...

Share on UMAP 2018 Intelligent User-Adapted Interfaces: Design and Multi-Modal Evaluation (IUadaptMe) Workshop Chairs' Welcome &Organization Authors: Ilknur Celik Cyprus International University, NICOSIA, CyprusView Profile , Ilaria Torre University of Genoa, GENOA, Italy ItalyView Frosina Koceva Christine Bauer Johannes Kepler Linz, LINZ, Austria AustriaView Eva Zangerle Universität Innsbruck, Bart Knijnenburg Clemson Clemson, SC, USA USAView Authors Info & Claims '18: Adjunct Publication...

The Wikidata platform is a crowdsourced, structured knowledgebase aiming to provide integrated, free and language-agnostic facts which are---amongst others---used by Wikipedias. Users who actively enter, review revise data on are assisted property suggesting system provides users with properties that might also be applicable given item. We argue evaluating subsequently improving this recommendation mechanism hence, assisting users, can directly contribute an even more consistent extensive...