Detecting Social Bots on Facebook in an Information Veracity Context
Details the algorithm which powers the BotBeGone platform. ICWSM 2019.
read hereA web application designed to allow social media users to regain control over the content that they consume. Scans pages and user feeds for malicious content, and provides tools for deletion and blocking. Bot detection is done using a novel machine learning algorithm. Developed with both node.js and Python (Flask) for the backend, HTML5/CSS/JS for the frontend.
Check it outA dataset created to aid the detection of misinformation and bots on social media platforms. Comprised mostly of Facebook comments, making it the first of its kind. Also includes web content from several news pages. Created using the Facebook Graph API and Python (particularly BeautifulSoup).
See codeA novel named-entity recognition system focusing on detecting rare and sparsely occuring named entities, powered by flexible context features of word forms. The model used statistics and NLP, lending itself well to classic machine learning.
See codeDetails the algorithm which powers the BotBeGone platform. ICWSM 2019.
read hereGoes into a lot of depth about the BuzzFace dataset. ICWSM 2018.
Read hereDetails a Named Entity Recognition algorithm. EMNLP 2017.
Read hereA correction to the Zipf-Mandelbrot law for word distributions in natural text.
Read here