| KEYWORD |
Debugging IF-THEN rules with a recommender system in the loop
keywords END-USER PROGRAMMING, HUMAN-COMPUTER INTERACTION, INTERNET OF THINGS, IOT, SMART HOME
Reference persons FULVIO CORNO, LUIGI DE RUSSIS, ALBERTO MONGE ROFFARELLO
Research Groups DAUIN - GR-10 - Intelligent and Interactive Systems - e-LITE
Thesis type EXPERIMENTAL, RESEARCH
Description Nowadays, end users can personalize their Internet of Things (IoT) ecosystem via visual programming platforms such as IFTTT or Zapier. In such platforms, users can program the joint behavior of their smart devices and online services through the definition of trigger-action programs, i.e., rules like "if the camera in the kitchen detects a movement, then send me a Telegram message." Despite apparent simplicity, trigger-action programming is often a complex task for non-programmers, as it is vulnerable to reasoning errors and conflicting rules. To solve these problems, end-users might use some debugging mechanisms placed on top of the previously mentioned platforms, e.g., https://elite.polito.it/research/research-topics/488-eudebug. However, fixing the problematic rules is something that it is totally delegated to the user, who can introduce new errors and conflicts.
The goal of this thesis is to explore novel debugging mechanisms that can suggest, through a recommender system, how to fix possible run-time problems in trigger-action programs. Recommendation approaches have been recently explored in the field of trigger-action programming, e.g., to support the composition of trigger-action rules or to translate, through a conversational approach, the user intentions into specific rules. Here, the idea is to combine recommendation approaches with debugging mechanisms: when a tool detects that a trigger-action program may generate a problem, a recommender system might be employed to suggest the user to change specific rules, triggers, and/or actions with proper alternatives that a) satisfy users’ preferences and previous behaviors, and b) fix the identified problem.
The thesis work can start from the extensive work on trigger-action rules definition, debugging, and recommendation of the research group. The outcomes of the work, if satisfying and appropriate, will be made freely available as an open source project.
Required skills Programmazione web. Conoscenze su sistemi di raccomandazione sono un plus.
Deadline 29/09/2022
PROPONI LA TUA CANDIDATURA