According to BERTopic, what do Danish Parties Debate on when they Address Energy and Environment?
Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › fagfællebedømt
Standard
According to BERTopic, what do Danish Parties Debate on when they Address Energy and Environment? / Navarretta, Costanza; Hansen, Dorte Haltrup.
Proceedings of the 3rd Workshop on Computational Linguistics for the Political and Social Sciences. Ingolstad : Association for Computational Linguistics, 2023. s. 59-68.Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › fagfællebedømt
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - According to BERTopic, what do Danish Parties Debate on when they Address Energy and Environment?
AU - Navarretta, Costanza
AU - Hansen, Dorte Haltrup
PY - 2023
Y1 - 2023
N2 - This paper investigates how two policy areas, Environment and Energy were dealt with by seven Danish left and right wing parties in their electoral manifestos (2007-2019) and parliamentary debates between 2009 and 2020. We analyse the two datasets both quantitatively and contentwise using BERTopic (Grootendorst, 2022) to generate subtopics about the two policy areas from the debates, and comparing the results with the content of the manifestos. BERTopic is trained with both multilingual and Danish word embeddings (https://certainly.io/blog/danish-bert-model/). BERTopic generated useful subtopics with both word embeddings, but the multimodal ones produced the largest number of relevant subtopics. Our study shows that only few parties have a consistent behavior in their electoral manifestos and parliamentary debateswith respect to the topics that they address. This can be explained in terms of issue and party competition as well as coalition and role of the parties as proposed by political science researchers.
AB - This paper investigates how two policy areas, Environment and Energy were dealt with by seven Danish left and right wing parties in their electoral manifestos (2007-2019) and parliamentary debates between 2009 and 2020. We analyse the two datasets both quantitatively and contentwise using BERTopic (Grootendorst, 2022) to generate subtopics about the two policy areas from the debates, and comparing the results with the content of the manifestos. BERTopic is trained with both multilingual and Danish word embeddings (https://certainly.io/blog/danish-bert-model/). BERTopic generated useful subtopics with both word embeddings, but the multimodal ones produced the largest number of relevant subtopics. Our study shows that only few parties have a consistent behavior in their electoral manifestos and parliamentary debateswith respect to the topics that they address. This can be explained in terms of issue and party competition as well as coalition and role of the parties as proposed by political science researchers.
M3 - Article in proceedings
SN - 979-8-89176-032-5
SP - 59
EP - 68
BT - Proceedings of the 3rd Workshop on Computational Linguistics for the Political and Social Sciences
PB - Association for Computational Linguistics
CY - Ingolstad
ER -
ID: 378760384