Abstract

NewsREEL Multimedia premiers 2018 as part of the MediaEval Benchmarking Initiative. The NewsREEL task combines recommendation algorithms with image and text analysis. Participants must predict engagement with news items based on text snippets and annotated Images. Several major German news portals have supplied data. The algorithms are evaluated in terms of precision on unknown data. This paper describes the task and the provided data in detail and explains the applied evaluation approach. The algorithms are evaluated based on Precision and Average-Precision for the top news items.

@inproceedings{...,
author = {Andreas Lommatzsch and Benjamin Kille and Frank Hopfgartner, Leif Ramming},
title = {MediaEval 2018 - Overview on NewsREEL Multimedia},
booktitle = {Proceedings of the MediaEval Benchmarking Initiative for Multimedia Evaluation 2018},
year = {2018},
issn = {1613-0073},
url = {http://ceur-ws.org/Vol-2283/},
location = {Sophia Antipolis, France},
publisher = {CEUR Workshop Proceedings},
}