HackHate | Devpost

result_page
front_page

Inspiration

The last elections have shown that politicians are getting away with posting hate speech and false information on their social media feeds.

What it does

We want to provide a tool that allows to analyse politicians social media feeds in regard to hate, information credibility and sentiment analysis. Moreover, it's not only a search page, but also a chrome plugin to directly extract and check content from websites.

How we built it

We have used several tools:

main language is python, pandas and nltk for sentiment analysis
the application is served with django
hatebase database for detecting hate speeches and words
our own database and duck duck go search for checking content credibility
javascript for interactive content on the webpage and for the chrome extension
twitter, facebook and reddit api for retrieving content
wikidata and sparql for fetching information about politicians

Challenges we ran into

Facts checking is tremendously hard
website credibility
hate speech analysis is not just single words checking
wikidata querying is not easy and can take a while
parsing and scrapping websites and search engines results
Not enough time to work on our own classifier, even though we have annotated data

Accomplishments that we're proud of

Good looking web application that works
the chrome extension is simple but effective
working hate speech detection

What we learned

generating and using apis
using and combining multiple content sources
generating meaningful visualizations
building chrome extension

What's next for HackHate

Getting a large dataset from reddit data and training our own machine learning classifier
Extending our database for credibility checking
annotating the web pages with chrome extension to give live credibility checking