A semi-supervised approach to classifying political agenda issues
| العنوان: | A semi-supervised approach to classifying political agenda issues |
|---|---|
| المؤلفون: | Tim Kreutz, Walter Daelemans |
| المصدر: | University of Antwerp Proceedings of the 1st Workshop on Computational Linguistics for Political Text Analysis (CPSS-2021) |
| مصطلحات موضوعية: | Computer. Automation, Linguistics |
| الوصف: | This paper presents a semi-supervised approach to classifying political texts with the Comparative Agendas Project coding scheme. Starting with limited domain knowledge in the form of ten seed words that are central to the meaning of a topic, new candidate textual indicators are found using a graph propagation algorithm over a semantic network of words and phrases. We show that there is a balance between precision and recall when it comes to the number of candidates to add to a lexicon for each topic, and optimize this balance on the basis of a development dataset. The automatically generated lexica substantially outperform the handmade CAP-lexicon in four tested genres: political party manifestos, news articles, parliamentary documents and social media texts. Besides having better discriminatory qualities, these lexica require less resources to generate and are more genre-independent |
| URL الوصول: | https://explore.openaire.eu/search/publication?articleId=dedup_wf_001::372b470c8df191eb2ddf4f22133293a4 https://hdl.handle.net/10067/1852840151162165141 |
| حقوق: | CLOSED |
| رقم الأكسشن: | edsair.dedup.wf.001..372b470c8df191eb2ddf4f22133293a4 |
| قاعدة البيانات: | OpenAIRE |
| الوصف غير متاح. |