Candidate Page example 2
Choosing the race to search candidates for
Choosing a candidate to research
Candidate Page example 1
Summary of candidates
Adam Berinsky, professor of political science at MIT, has surmised that making voting easier does little to increase voter turnout (https://ssir.org/increasing_voter_turnout/entry/making_voting_easier_doesnt_increase_turnout), citing policies like the Motor Voter Act’s failure to increase turnout rates consistently. Instead, his and other research suggests that we need to think of the cognitive “costs” of voting, because our level of political engagement and interest direct our political behavior. With this in mind, we wanted to search for a way to easier and more efficient to both display and access data about local, state, and national elections and in particular various candidate’s policy positions.
What it does
Our site automatically pulls and synthesizes the abundance of public but extremely noisy data available on internet (e.g. internet, government hosted sites, blogs) into a concise summary of a candidate's stance on key issues. Users enter their zip code to find the elections they can vote in and the candidates running for office in these elections. The site automatically generates pages and descriptions for each candidate. Each topic card contains a summary of all the articles found on the internet about the candidate’s stance on a certain issue, as shown below.
How we built it
Backend Zipcode querying , Django
For the text summarization portion of our project, we used a __ . We do this through seq2seq + bert (abstractive summarization)
We do this using the algorithm introduced by this paper, specifically by combining Bidirectional Encoder Representations from Transformers (or BERT) and a combination of extractive and abstractive summarization models.
Other optimizations introduced by this paper.
Frontend HTML, CSS, Bootstrap, JQuery
Challenges we ran into
import statements bad documentation on the part of other people finding a good model rigorous AB testing debugging Determining data types of API returns
Accomplishments that we're proud of
- Getting the summarization to finally work, although the first summary of an article we got was “michael moore me me me”.
- Deciding on a final website design
What we learned
-Communication is crucial to a good workflow -Raleway is a good font -Teamwork makes the dream work
What's next for Political Synthesis
-Improve text summarizer -More elegant UI/UX -Expand to different local elections