If you enjoy mining, analysing, classifying, or learning from either a mass of data, or perhaps an online network, this track may give you interesting research to do. Pick any data-rich open problem which needs a mining or machine-learning algorithm, any open problem from network or Web science, or any computationally hard problem which needs an optimization algorithm from the field of Artificial Intelligence (AI).
You may start with data from social media (StackExchange, Twitter, Wikipedia, Reddit), online data describing products that our society likes and consumes (song features and charts, Amazon products on sale), logistics data (airplane or ship telemetry, readings from road sensors), or a Web crawl. You can get such datasets either from our existing collection, or from data archives on the Internet (https://www.kaggle.com/, https://archive.org/), or even scrape it yourselves from the Web or with a social network's API.
You may then apply or develop: mining algorithms, neural networks, classifiers, natural-language processing, information retrieval, graph mining, evolutionary algorithms, heuristics, or any combination of such algorithms, to solve a research question of your choice. You can get inspiration from past editions of this track (see for example last year's edition, http://referaat.cs.utwente.nl/conference/27/paper), or simply talk to the track chair.
Here is a list of current research questions, for inspiration:
Can I mine knowledge about society from people's behaviour online? For example, (1) the books people co-buy on Amazon can show which political leanings correlate with preferences for which science (do left-wing politics go with reading computer science topics?), (2) the way tweets propagate in the network of Twitter followers can tell which type of tweet content is viral in an online social network, and which not, (3) the way people now form specialised communities online can tell which interests correlate (do right-wing politics go with entrepreneurship?), (4) the questions asked on StackOverflow can tell which programming frameworks are learnt or used together by programmers.
Can I help improve people's safety online? This could be done by coming up with methods to detect bots, malware, verbal abuse, or the spread of fake news, in any online social media.
Can I help make existing AI algorithms more self-explanatory to their users? (This is the field of eXplainable AI, or XAI). This could be done by (1) stress-testing your AI algorithm of choice to see how sensitive its classification is to small modifications in the input data (see http://www.evolvingai.org/fooling), (2) measuring how biased that algorithm is (see https://algotransparency.org), (3) adding means for the AI algorithm to "explain" why it reached a certain conclusion, in simple, human-readable terms ("I turned the lights on in your office because...").
Can I improve online search engines, or recommender algorithms in social networks?
For further information on the content of this track, you may contact the track chair Doina Bucur.