Naslov
Combining Sentiment Analysis with Linguistics: Figurative Speech Detection in Croatian Online Text

Autori
Ptiček, Martina

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, ostalo

Izvornik
Proceedings of the 11th International Doctoral Seminar (IDS 2019) / Vrček, Neven ; Moravčik, Oliver - Varaždin : Fakultet organizacije i informatike Sveučilišta u Zagrebu, 2019, 29-30

ISBN
978-953-6071-67-8

Skup
30th Central European Conference on Information and Intelligent Systems (CECIIS 2019)

Mjesto i datum
Varaždin, Hrvatska, 02.10.2019. - 04.10.2019

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
web text analysis, figurative speech, figurative language processing, irony, sentiment analysis

Sažetak
Figurative speech (i.e. language figures such as irony, sarcasm, metaphor etc.) makes an important part of communication including the online communication and texts, such as posts, reviews, tweets etc. [1, 240]. In order to properly analyse the sentiment of texts, attention should be given to the fact that their meaning sometimes can be “hidden” due to the use of figurative language. Knowing that recognition of figurative language can be a challenge even for a human being, analysing and recognizing figurative language is an especially challenging task in natural language processing and in machine learning. The analysis has to be twofold and in first instance encompass decoding of the literal meaning after which the second code can be decoded that will lead to identification of the “real” meaning.To develop a model of a sentiment analysis that will take into the account figurative language, this doctoral research will start from analysing irony, as the common figure in the use of figurative language. Irony will be analysed from the linguistical perspective with the goal to identify important aspects which will allow us to discriminate ironic from non-ironic sentence. After these characteristics are defined, a model will be developed which will be able to discern between the ironic and non-ironic sentence. Developed model will be tested in conducting the task of sentiment analysis this test will be used to support the hypothesis that sentiment analysis will be more successful if the model takes into account that figurative speech is present, compared to the sentiment analysis that does not recognize the figurative speech. The dataset that will be used for this experiment will consist of user-generated content published on Croatian forums and comments on selected news portals. It is expected that this model won’t give the same results in different domains (e.g. politics, sport, fashion etc.) therefore its relevance will be tested for each of the selected domains, which will enable a comparative insight into the applicable value of the model in specific domains.

Izvorni jezik
Engleski

Znanstvena područja
Informacijske i komunikacijske znanosti

POVEZANOST RADA