Demos & Tools.
Demos
OPIAS: Over-Personalization in Information Access Systems
Bursting the Bubble: Social Media Demo
This demo presents the user with a social media homepage with posts about particular topics (e.g. Nuclear Energy, GMOs) and asks the user to “like” posts or comments that they agree with. Upon interacting with a post/comment, more posts that match that topic and point of view (e.g. supporting Nuclear Energy) are loaded, in a feedback loop that leaves out other points of view and other topics.
Bursting the Bubble: Search Engine Demo
This demo gives the user the option to search for a particular topic (out of a list of topics available), then upon seeing the results, the option to search for the same topic in other manners or with another ‘account’. So, the user can see how the search results change based on the language or wording of the query (e.g. Kıbrıs vs. Κύπρος, scientist vs. woman scientist) as well as the impact of the user’s profile (e.g. age, gender, location) on the results chosen.
OpenTag: Understanding Human Perceptions of Image Tagging Algorithms
An award-winning (HCOMP 2020 – Best Demo Award) platform that fuses the auditing process with a crowdsourcing approach. Users can upload an image, which is then analyzed by various Image Tagging Algorithms (ITAs), resulting in multiple sets of descriptive tags. With OpenTag, the user can observe and compare the output of multiple ITAs simultaneously, while researchers can study the manner in which users perceive this output. Finally, using the collected data, further audits can be performed on ITAs.
Tools
CovNeTiM: Covid News Topic Modeler
A python toolbox that consists of NLP techniques for extracting thematic topics per news source. The tool was created during the first wave of the COVID-19 pandemic for analyzing the COVID-related news and updates. The tool used four well-known news sources to analyze their published articles such as The New York Times (US), The Guardian (UK), ANSA (Italy), and Cyprus Mail (Cyprus). Using this tool’s functionality one can map the current talk around COVID-related articles and extract themes and useful semantic tags based on the specialized taxonomy created for the tool based on regular expressions.
CFD background stimuli generation
This code takes input images from the Chicago Face Database (which must have had their backgrounds rendered transparent from the original white). It then combines them (in all possible ways) with some pre-set backgrounds to create stimuli that have a person in the foreground and a specific environment in the background.
ITAAuditToolbox
An audit toolbox for analyzing Image Tagging Algorithms (ITAs) output and behavior. (TBA)