Napredna pretraga

Pregled bibliografske jedinice broj: 599328

Identifying Fear Related Content in Croatian Texts


Načinović, Lucia; Perak, Benedikt; Meštrović, Ana; Martinčić-Ipšić, Sanda
Identifying Fear Related Content in Croatian Texts // Zbornik Osme konference Jezikovne tehnologije, 8. do 12. oktober 2012, [Ljubljana, Slovenia] : zbornik 15. mednarodne multikonference Informacijska družba - IS 2012, zvezek C = Proceedings of the Eighth Language Technologies Conference, October 8th-12th, 2012, Ljubljana, Slovenia : proceedings of the 15th International Multiconference Information Society - IS 2012, volume C / Tomaž Erjavec, Jerneja Žganec Gros (ur.).
Ljubljana: Institut »Jožef Stefan«, Ljubljana, 2012. str. 153-156 (predavanje, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)


Naslov
Identifying Fear Related Content in Croatian Texts

Autori
Načinović, Lucia ; Perak, Benedikt ; Meštrović, Ana ; Martinčić-Ipšić, Sanda

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
Zbornik Osme konference Jezikovne tehnologije, 8. do 12. oktober 2012, [Ljubljana, Slovenia] : zbornik 15. mednarodne multikonference Informacijska družba - IS 2012, zvezek C = Proceedings of the Eighth Language Technologies Conference, October 8th-12th, 2012, Ljubljana, Slovenia : proceedings of the 15th International Multiconference Information Society - IS 2012, volume C / Tomaž Erjavec, Jerneja Žganec Gros - Ljubljana : Institut »Jožef Stefan«, Ljubljana, 2012, 153-156

ISBN
978-961-264-048-4

Skup
Jezikovne tehnologije

Mjesto i datum
Ljubljana, Slovenija, 8-9.10.2012

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Semantic analysis; fear; identification of emotion; metonymic constructions

Sažetak
This paper presents the initial work for the task of identifying the emotion of FEAR in texts written in the Croatian language. For the purpose of this analysis, text articles, blogs and online comments were collected from specific Croatian websites and classified into two categories: “Fear present” and “Fear not present”. In the process of classification, supervised machine learning method based on Naïve Bayes algorithm was used. We experimented with different sets of features to compare their impact on the accuracies of the learnt classifiers and to determine which set of features yields better result (accuracy). The first set of features was constructed on a semantic model of embodied metonymic domains of fear derived from the cognitive semantic study of metonymic constructions of fear in Croatian. The second set of features contained only lexical concepts of fear, its synonyms and direct hyponyms. The third set is a combination of both sets. The results are presented in terms of accuracies, indicating that both embodied metonymic and lexically expressed features can be relevant for the emotion identification in texts.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo, Informacijske i komunikacijske znanosti, Filologija



POVEZANOST RADA


Projekt / tema
318-0361935-0852 - Govorne tehnologije (Ivo Ipšić, )

Ustanove
Filozofski fakultet, Rijeka,
Sveučilište u Rijeci - Odjel za informatiku