Short biography


Since April 2018, I am the CTO at Accurat.ai. We offer solutions to better understand the offline behaviour of app users.

I worked as a Data Scientist at Realo, a fast growing real estate startup in Europe, between October 2016 and March 2018. After two months, I have been promoted to Lead Data Scientist. My main responsibilities were implementing and improving machine learning pipelines for the estate price estimates, classification tasks and recommendations.

I did a summer Research Internship at the Yahoo! Advertising Science team in London (July-Sept 2016). The focus of my internship was to optimize mobile advertising by targeting the right audience during major events (e.g. major sport events, Christmas, political elections) using big data technologies.

Between August 2011 and June 2016, I worked as a PhD student at the Department of Information Technology (INTEC) of the Faculty of Engineering at Ghent University. I was funded by a PhD grant of the Agency for Innovation by Science and Technology (IWT). My research included (geographic) information retrieval and data mining. Specifically I constructed scalable strategies to extract structured place, event and sentiment information from social media. In addition, I developed methodologies to monitor, model and predict the viral nature of online news on social media.

Before my PhD, I obtained with great distinction a Master degree in Computer Science Engineering (Ghent University, 2011, burgerlijk ingenieur), with a major degree in Software Engineering and a minor degree in Business Science.

Contact


Steven Van Canneyt
CTO - Data Scientist
Accurat
Hundelgemsesteenweg 316A
9820 Merelbeke
Belgium

Mobile: +32 (0)495 42 47 14

LinkedIn: Steven Van Canneyt

PhD Dissertation


Knowledge Extraction and Popularity Modeling Using Social Media, October 2016, Ghent University, Belgium (pdf).

Publications

  1. Describing patterns and disruptions in large scale mobile app usage data (pdf)
    Steven Van Canneyt, Marc Bron, Andy Haines, Mounia Lalmas
    In the proceedings of the 26th International Conference on World Wide Web Companion, pages 1579-1584, 2017.

  2. Modeling and predicting the popularity of online news based on temporal and content-related features (published - camera ready)
    Steven Van Canneyt, Philip Leroux, Bart Dhoedt, Thomas Demeester
    Multimedia Tools and Applications, pages 1-28, 2017.

  3. Representation learning for very short texts using weighted word embedding aggregation (pdf)
    Cedric De Boom, Steven Van Canneyt, Thomas Demeester, Bart Dhoedt
    Pattern Recognition Letters, 80, pages 150-156, 2016.

  4. Categorizing events using spatio-temporal and user features from Flickr (published - camera ready)
    Steven Van Canneyt, Steven Schockaert, Bart Dhoedt
    Information Sciences, 328, pages 76-96, 2016.

  5. Learning representations for tweets through word embeddings (pdf)
    Cedric De Boom, Steven Van Canneyt, Thomas Demeester, Bart Dhoedt
    In the proceedings of the Belgian-Dutch Conference on Machine Learning, pages 60-62 , 2016.

  6. Categorizing events using spatio-temporal and user features from Flickr (abstract) (pdf)
    Steven Van Canneyt, Steven Schockaert, Bart Dhoedt
    In the proceedings of the 14th Dutch-Belgian Information Retrieval Workshop (DIR), page 14, 2015.

  7. Learning semantic similarity for very short texts (pdf)
    Cedric De Boom, Steven Van Canneyt, Steven Bohez, Thomas Demeester, Bart Dhoedt
    In the proceedings of the 2nd ICDM International Workshop on Representation Learning for Semantic Data, pages 1229-1234, 2015.

  8. Optimizing the popularity of Twitter messages through user categories (pdf)
    Rupert Lemahieu, Steven Van Canneyt, Cedric De Boom, Bart Dhoedt
    In the proceedings of the 2nd ICDM International Workshop on Social Multimedia Data Mining, pages 1396-1401, 2015.

  9. Semantics-driven event clustering in Twitter feeds (best paper award) (pdf)
    Cedric De Boom, Steven Van Canneyt, Bart Dhoedt
    In the proceedings of the 5th WWW International Workshop on Making Sense of Microposts, pages 2-9, 2015.

  10. Topic-dependent sentiment classification on Twitter (pdf)
    Steven Van Canneyt, Nathan Claeys, Bart Dhoedt
    In the proceedings of the 37th European Conference on Information Retrieval (ECIR), pages 441-446, 2015.

  11. Estimating the semantic type of events using location features from Flickr (pdf)
    Steven Van Canneyt, Steven Schockaert, Bart Dhoedt
    In the proceedings of the 8th ACM SIGSPATIAL International Workshop on Geographic Information Retrieval, pages 57-64, 2014.

  12. Detecting newsworthy topics in Twitter (pdf)
    Steven Van Canneyt, Matthias Feys, Steven Schockaert, Thomas Demeester, Chris Develder, Bart Dhoedt
    In the proceedings of the SNOW 2014 Data Challenge, pages 25-32, 2014.

  13. Discovering and characterizing places of interest using Flickr and Twitter (published - camera ready)
    Steven Van Canneyt, Steven Schockaert, Bart Dhoedt
    International Journal on Semantic Web and Information Systems (IJSWIS), 9(3), pages 77-104, 2013.

  14. Using social media to find places of interest: A case study (best paper award) (pdf)
    Steven Van Canneyt, Steven Schockaert, Olivier Van Laere, Bart Dhoedt
    In the proceedings of the 1st ACM SIGSPATIAL International Workshop on Crowdsourced and Volunteered Geographic Information, pages 2-8, 2012.

  15. Detecting places of interest using social media (pdf)
    Steven Van Canneyt, Steven Schockaert, Olivier Van Laere, Bart Dhoedt
    In the proceedings of the 2012 IEEE/WIC/ACM International Conference on Web Intelligence, pages 447-451, 2012.

  16. A context-aware tourism recommendation system (shortlisted for best poster awards) (pdf)
    Steven Van Canneyt, Bart Dhoedt
    In the proceedings of the 12th UGent-FEA PhD symposium, page 61, 2011.

  17. Time-dependent recommendation of tourist attractions using Flickr (pdf)
    Steven Van Canneyt, Steven Schockaert, Olivier Van Laere, Bart Dhoedt
    In the proceedings of the 23rd Benelux Conference on Artificial Intelligence (BNAIC), pages 255-262, 2011.

Projects

  1. PROVIDENCE: Predicting the online virality of entertainment and news content (more information)
    Steven Van Canneyt, Thomas Demeester, Philip Leroux, Thomas Vanhove
    The PROVIDENCE research project aims to optimize online news publication strategies by anticipating the predicted viral nature of news on social media, and is carried out in collaboration with Flemish media companies. I developed a large-scale monitoring framework to track the popularity of news items and their interactions on social media in real time. I used the monitored data to construct novel methodologies which analyze and predict the news consumption and news sharing behavior by users.
    The project partners are IBCN - Gent University, SMIT - VUB, VRT, NewsMonkey, Massive Media Match and iMinds Media Innovation Centrum, the project is funded by IWT.

Invited talks

  1. Discover the value of your house: The Realo Estimate approach.
    Steven Van Canneyt
    For the real estate students of the University College Ghent, 2017.

  2. Increase your popularity on social media by using the right hashtags.
    Steven Van Canneyt
    At the 'Friday meeting' of the VRT Research and Innovation department, 2014.

  3. Analysis of tourist behaviour using social media.
    Steven Van Canneyt
    At the Waterways for Growth meeting, 2012.

Heat Map


Regions in Ghent (Belgium) related to a given place type, based on the tags of Flickr photos. For more information, see

Detecting places of interest using social media (pdf)
Steven Van Canneyt, Steven Schockaert, Olivier Van Laere, Bart Dhoedt
In the proceedings of the 2012 IEEE/WIC/ACM International Conference on Web Intelligence, pages 447-451, 2012.

place type: