Boshko Koloski's repositories
Multilingual-Detection-of-Fake-News-Spreaders-via-Sparse-Matrix-Factorization
Fake news is an emerging problem in online news and social media. Efficient detection of fake news spreaders and spurious accounts across multiple languages is becoming an interesting research problem, and is the key focus of this paper. Our proposed solution to PAN 2020 fake news spreaders challenge models the accounts responsible for spreading the fake news by accounting for different types of textual features, decomposed via sparse matrix factorization, to obtain easy-to-learn-from, compact representations, including the information from multiple languages. The key contribution of this work is the exploration of how powerful and scalable matrix factorization-based classification can be in a multilingual setting, where the learner is presented with the data from multiple languages simultaneously. Finally, we explore the joint latent space, where patterns from individual languages are maintained. The proposed approach scored second on the 2020 PAN shared task for identification of fake news spreaders.
Know-your-Neighbors-Efficient-Author-Profiling-via-Follower-Tweets
User profiling based on social media data is becoming an increasingly relevant task with applications in advertising, forensics, literary studies and sociolinguistic research. Even though profiling of users based on their textual data is possible, social media such as Twitter offer also insight into the data of a given user’s followers. The purpose of this work was to explore how such follower data can be used for profiling a given user, what are its limitations and whether performances, similar to the ones observed when considering a given user’s data directly can be achieved. In this work we present our approach, capable of extracting various feature types and, via sparse matrix factorization, learn a dense, low-dimensional representations of individual persons solely from their followers’ tweet streams. The proposed approach scored second in the PAN 2020 Celebrity profiling shared task, and is computationally non-demanding.
covid19-fake-news
Covid19 Fake News
dh2021
DH2021
graph2gauss
Gaussian node embeddings. Implementation of "Deep Gaussian Embedding of Graphs: Unsupervised Inductive Learning via Ranking".
PAN2020-Celebrity-Profiling
Know-your-Neighbors-Efficient-Author-Profiling-via-Follower-Tweets
rakun
Rank-based Unsupervised Keyword Extraction via Metavertex Aggregation
rakun2
RaKUn 2.0 - A fast keyword detection algorithm