Master thesis opportunities
Our guest lecturers offer a number of exciting Master thesis opportunities. If you are interested, please contact the person listed as contact point for each project!
Detecting Fake Users on Twitter
Magnus Rosell, magnus.rosell@foi.se
Twitter is a social media service that have got an increasing number of users in recent years, However, there is also an increasing number of people who are misusing it in various ways. It is therefore of interest for Twitter as well as organizations who are interested in analyzing social media to be able to detect users who in some sense are "fake". Fake users include
- bots that among other things are used to send spam
- real user accounts that are hijacked in order to spread disinformation
- users that are bought to write certain things
Various types of metadata or features can potentially be used to capture such "fake" users but it is far from obvious which features that work best or how to combine them. In order to detect bots it may be useful to look at features such as when the user account was created, how many followers the user has, whether or not there is a profile picture, etc. Changes in user behavior are probably more useful for detecting hijacked Twitter accounts, i.e., sudden deviations in features such as the amount of tweets, the content of tweets, etc.
Although the exact scope of the thesis can be discussed, we would like to work with students who have previous experience in machine learning, anomaly detection, and/or text analysis. The student should be able to conduct a review of existing literature in order to identify current state-of-the-art algorithms, implement a subset of these, and evaluate the implemented algorithms on appropriate test datasets.
Exjobb i språkteknologi
Jussi Karlgren, jussi@kth.se
http://www.csc.kth.se/tcs/humanlang/exjobb.html