The conference on Human Computation and Crowdsourcing (HCOMP2018) was recently held in Zurich, Switzerland. The conference is unique for its interdisciplinary focus, where themes range from Artificial Intelligence (AI), Human Computer Interaction (HCI), to crowdsourcing and digital humanities. At this conference QROWD were thrilled to present some of their original and insightful findings in smart mobility in a number of different ways.
Firstly, the QROWD team were privileged to announce that the paper “How Biased is Your NLG Evaluation” won the Best Paper award at the #CrowdBias workshop during the conference. The paper discusses issues with human assessments by crowdworkers and experts, used for the evaluation of systems for text generative tasks. The paper generates a list of summaries from DBpedia triples using state of the art neural network architecture.
You can read a copy of the full paper here: https://www.alessandrochecco.info/static/CrowdBias2018_paper_8.pdf
Secondly, the project presented a poster on the various features and initiatives of the project. In this poster we explain what we do, why it is important, and what we can offer individuals. Furthermore, we feature three core solutions developed by the project which aim to solve some of problems with smart mobility. These are:
Virtual city explorer – This is a crowdsourcing solution for mobility infrastructure maps. Workers can login to the city virtual viewer and follow instructions to locate static items, for example bike racks. Individuals can be confined into certain areas of the city, and can see whether other bike racks have previously been logged on the map. Previously logged items are flagged as “taboo”, which allows the system to optimise cost and coverage as workers can only log non-taboo items: meaning that individuals are encouraged to continually find new items.
QROWDsmith – QROWDsmith is a standalone crowdsourcing platform that aims to improve the engagement of contributors or crowdworkers. It does this through two main ways. Firstly, it increases productivity by using gamification methods. These are in the form of leaderboards, unlocking badges for new tasks, and logging personal scores. Secondly, it allows individuals to compete in tasks, as well as offering cooperative tasks to optimise productivity.
Citizen-in-the-loop – This feature confirms ML predictions on data generated by citizens that use mobility services across a city. It asks citizens to confirm their travel methods, time taken, and other metadata about their commutes within the city. This simultaneously generates more reliable data, as well as engaging citizens more with mobility.
Finally, the project presented and demonstrated some of these features. These will be showcased for the wider public in the upcoming weeks. We look forward to sharing these with you in the near future focusing on each of these unique solutions in more detail. Stay tuned!