Indexes
Media Depolarization Index
The methodology employed in the ISET Policy Institute's Media Polarization Index relies heavily on two primary Natural Language Processing (NLP) models: "Word2Vec" and its extension, "Doc2Vec". The authors trained a Georgian language "Doc2Vec" model specifically to capture semantic meanings in Georg
ian political news articles. This model was trained on a corpus exceeding 250,000 online political news articles gathered from diverse sources. Following training, the model is applied to political news articles from popular media outlets (“Imedi”, “Mtavari”, “TV Pirveli”, “1TV” (Public Broadcaster), “Formula”, “PosTV” and “Rustavi2”). The vectors generated by these models exist in a high-dimensional space and dissimilarity among news sources is measured using cosine similarity metrics. The politically biased dissimilarity between media platforms is calculated as the difference between the total dissimilarity and the average total dissimilarity within clusters (the research identifies two media clusters). The Media Polarization Index is a weighted average of political dissimilarities between media outlets, where weights are proportional to their ratings.
Filter By:
Date From
Date To
April 2024 | Media (de)Polarization Index
30
April
2024
The Index shows that media polarization increased in April 2024 which could be driven by reintroducing the so-called "Russian Law".
Georgia Media (de)Polarization Index
20
March
2024
The Index shows an increase in media polarization since 2020, particularly acute since early 2022. While the Index captures significant polarization around specific events like elections, its response varies across different events and developments.