http://hdl.handle.net/10379/5420 2017-10-29T22:46:12Z 2017-10-29T22:46:12Z Ihsan, Ullah, Lane, Caoilfhionn Drury, Brett Mellotte, Marc Madden, Michael G. http://hdl.handle.net/10379/6830 2017-09-27T01:02:00Z 2017-07-20T00:00:00Z

Open social data crime analytics Ihsan, Ullah,; Lane, Caoilfhionn; Drury, Brett; Mellotte, Marc; Madden, Michael G. Crime is under-reported. Reporting crime requires the victim to complete a number of administrative obligations. These obligations, as well as the nature of the crime, may create an inertia that discourages the reporting of the crime (for example, being defrauded might damage a financial organisation s reputation). However, there may be information leaks from compromised organizations, via affected customers on social media. A key advantage of using social data is that it is often immediate, and can have indications of the nature of a crime such as (1) named entities, for example, Bitcoin or PayPal; (2) geocoding information; and (3) the affected persons. Our aim in this work is to use social media platforms e.g. Twitter, Reddit, Facebook, etc. to detect signals of cybercrime incidents. Such signaling is arguably a better indicator of the extent and effect of cybercrime than traditional reporting methods.

2017-07-20T00:00:00Z Heitmann, Benjamin Hayes, Conor http://hdl.handle.net/10379/6561 2017-06-03T01:01:10Z 2015-06-01T00:00:00Z

The role of negative results for choosing an evaluation approach - a recommender systems case study Heitmann, Benjamin; Hayes, Conor We describe a case study, which shows how important negative results are in uncovering biased evaluation methodologies. Our re- search question is how to compare a recommender algorithm that uses an RDF graph to a recommendation algorithm that uses rating data. Our case study uses DBpedia 3.8 and the MovieLens 100k data set. We show that the most popular evaluation protocol in the recommender sys- tems literature is biased towards evaluating collaborative filtering (CF) algorithms, as it uses the rating prediction task. Based on the negative results of this first experiment, we find an alternative evaluation task, the top-k recommendation task. While this task is harder to perform, our positive results show that it is a much better fit, which is not biased to- wards either CF or our graph-based algorithm. The second set of results are statistically significant (Wilcoxon rank sum test, p

2015-06-01T00:00:00Z Silva, Vivian S. Hürliman, Manuela Davis, Brian Handschuh, Siegfried Freitas, André http://hdl.handle.net/10379/6244 2016-12-15T02:01:00Z 2016-12-12T00:00:00Z

Semantic relation classification: task formalisation and refinement Silva, Vivian S.; Hürliman, Manuela; Davis, Brian; Handschuh, Siegfried; Freitas, André The identification of semantic relations between terms within texts is a fundamental task in Natural Language Processing which can support applications requiring a lightweight semantic interpretation model. Currently, semantic relation classification concentrates on relations which are evaluated over open-domain data. This work provides a critique on the set of abstract relations used for semantic relation classification with regard to their ability to express relationships between terms which are found in a domain-specific corpora. Based on this analysis, this work proposes an alternative semantic relation model based on reusing and extending the set of abstract relations present in the DOLCE ontology. The resulting set of relations is well grounded, allows to capture a wide range of relations and could thus be used as a foundation for automatic classification of semantic relations.

2016-12-12T00:00:00Z Marsden, Mark Mohedano, Eva McGuinness, Kevin Calafell, Andrea Giro-i-Nieto, Xavier O'Connor, Noel E. Zhou, Jiang Azavedo, Lucas Daudert, Tobias Davis, Brian Hürlimann, Manuela Afli, Haithem Du, Jinhua Ganguly, Debasis Li, Wei Way, Andy Smeaton, Alan F. http://hdl.handle.net/10379/6242 2016-12-15T02:00:51Z 2016-11-14T00:00:00Z

Dublin City University and Partners' participation in the INS and VTT Tracks at TRECVid 2016 Marsden, Mark; Mohedano, Eva; McGuinness, Kevin; Calafell, Andrea; Giro-i-Nieto, Xavier; O'Connor, Noel E.; Zhou, Jiang; Azavedo, Lucas; Daudert, Tobias; Davis, Brian; Hürlimann, Manuela; Afli, Haithem; Du, Jinhua; Ganguly, Debasis; Li, Wei; Way, Andy; Smeaton, Alan F. Dublin City University participated with a consortium of colleagues from NUI Galway and Universitat Polit`ecnica de Catalunya in two tasks in TRECVid 2016, Instance Search (INS) and Video to Text (VTT). For the INS task we developed a framework consisting of face detection and representation and place detection and representation, with a user annotation of top-ranked videos. For the VTT task we ran 1,000 concept detectors from the VGG-16 deep CNN on 10 keyframes per video and submitted 4 runs for caption re-ranking, based on BM25, Fusion, Word2Vec and a fusion of baseline BM25 and Word2Vec. With the same pre-processing for caption generation we used an open source image-to-caption CNN-RNN toolkit NeuralTalk2 to generate a caption for each keyframe and combine them.

2016-11-14T00:00:00Z