What it does
We want to help people to find political information that is relevant for them and by doing so improve engagement in democracy.
It notifies the user when there are new events in the long dwindling political processes.
How we built it
We handcrafted a backend in Python utilizing gensim, sklearn, numpy and flask to implement a decision tree based classifier that maps unlabeled documents to tags/categories. It also stores the relationships between documents in a Neo4j graph database.
The application architecture is illustrated in the figure below.
Challenges we ran into
Understanding the data and making it possible to visualise it. The Riksdags data was very unstructured and thus hard to link documents even though the Riksdag has a very long standing history of using unique identifiers for all documents.
Accomplishments that we're proud of
We have achieved a lot of features and worked well together as a team. We were able to separate the responsibilities among the team members which means we were able to be productive.
What we learned
We as a team learned a lot about the structure of the open data delivered by the different parties. And we found some interesting data sets. The problem however was to find the right data and to interpret the data correctly. We improved a lot in understanding complex data sources.
What's next for this service:
- Open Source the code.
- Fix the document hierarchy tree.
- Focus on security, reliability & code quality.
- Do user research and decide on platform.