Information flow between news articles: Slovene media case study
We present results of a study on usage of text similarity measures based on co-occurrence of words and phrases to classify a relation between a pair of news articles (i.e. no relation, both based on a common source, one based on the other).
For each Slovenian article written in Slovene and published
online on 27th June 2016, we found the most similar release
from the Slovenian Press Agency (STA) database to obtain
a list of candidate article-source pairs. Four experts from
STA were asked to score the pairs, and their annotations
were used to train classifiers and evaluate their accuracy.