Version française

Transparency is good. Providing the keys to understanding datasets published in open access is better: this was the common thread that united the eight active teams during the Open Data x Democracy in Action hackathon, in which some thirty enthusiasts participated.

This year, the focus was on democracy, thanks to the many datasets provided by the Chamber of Deputies since 2022: more than twenty, ranging from parliamentary questions to votes in plenary sessions, not forgetting the list of petitions.

While this data is regularly updated and available in various file formats, it still needs to be attractive, easy to access and tailored to user needs. It is this aspect that the developers present at the event wanted to explore.

One of the projects takes the week's votes and, thanks to an LLM, generates a one-minute podcast. The idea is to target a young audience and raise awareness about the Chamber's work.

Still aiming to meet users' specific needs, a prototype, "my chamber," aims to offer personalised information based on each user's profile — for example, a list of petitions that might be relevant to him or her.

Who takes the time to regularly check the petitions submitted on petitiounen.lu? Here is "Petinder", an app billed as the Tinder of petitions, where users select petitions they consider worthy of interest. They then receive a list of petitions they are encouraged to sign.

Another approach focused on the MPs themselves, comparing their attendance at plenary sessions, their votes, and the number of bills they introduced. Some indicators were consolidated at the political party level.

This original mapping of votes links each MP's decision to his or her seat in the Chambre, which offers a new perspective on parliamentary activity, grouped by party and by session.

Artificial intelligence was also featured in the form of two chatbots where users can freely ask questions related to petitions, debates, and laws. The first AI, based on multilingual semantic search, provides a summary accompanied by diagrams. The second imports the data into an SQL database beforehand, and an LLM performs the queries, which limits the AI's erroneous statements.

Overall, the quality of the work was excellent. These contributions provide a wealth of ideas and scenarios to inspire the Chamber as it continues its Open Data adventure.