Jan Marco, Lets stop living a lie:
In the current media landscape, control over distribution has become almost as important as the actual creation of content, and that has given Facebook a huge amount of power
The impact that Facebook newsfeed has in the formation of opinions in the real world is so huge that it potentially affected the 2016 election decisions, however these newsfeed were not completely accurate.
Our solution? FiB because With 1.5 Billion Users, Every Single Tweak in an Algorithm Can Make a Change, and we dont stop at just one.
What it does:
Our algorithm is two fold, as follows:
Content-consumption: Our chrome-extension goes through your facebook feed in real time as you browse it and verifies the authenticity of posts. These posts can be status updates, images or links.
Our backend AI checks the facts within these posts and verifies them using image recognition, keyword extraction, and source verification and a twitter search to verify if a screenshot of a twitter update posted is authentic. The posts then are visually tagged on the top right corner in accordance with their trust score. If a post is found to be false, the AI tries to find the truth and shows it to you.
Content-creation: Each time a user posts/shares content, our chat bot uses a webhook to get a call. This chat bot then uses the same backend AI as content consumption to determine if the new post by the user contains any unverified information. If so, the user is notified and can choose to either take it down or let it exist.
How we built it:
Our chrome-extension is built using javascript that uses advanced web scraping techniques to extract links, posts, and images. This is then sent to an AI. The AI is a collection of API calls that we collectively process to produce a single “trust” factor.
The APIs include Microsoft’s cognitive services such as image analysis, text analysis, bing web search, Twitter’s search API and Google’s Safe Browsing API. The backend is written in Python and hosted on Heroku. The chatbot was built using Facebook’s wit.ai
Inderdaad is in de code ook te zien bij wie ze met de bij FB losgeweekte flarden eventuele onzin onder andere allemaal langsgaan voor hun waarheidsvinding:
HackPrincetonF16/backend/imageverify.py
api.projectoxford.ai api.mywot.com api.cognitive.microsoft.com gateway.watsonplatform.net
Dit stukje source blaft ook allerlei - ik neem aan, free-starter’ - api-keys voor toegang tot deze services door de spijlen van het tuinhekje.
Splunk wil maar niet op me groeien? (vgl. grow on me). Genoemde ‘AI’ van IBM Watson op jouw link loslaten:
Extracts people, companies, organizations, cities, geographic features, and other entities from your content, and optionally detects the sentiment of each entity.
Entity Relevance Sentiment Type ================================================ Splunk 0.926858 positive Company Security Experts 0.370114 positive JobTitle Utrecht 0.277183 neutral City
Sterkte morgen: afgaand op het door Watson gemelde ‘sentiment’, jij zal moe maar voldaan weer thuiskomen, geheel doorstraald van nieuwste inzichten op het gebied van dataslurpen o/o