Računalnolingvistička analiza korisničkih komentara na internetskim portalima

Matković, Antonia. (2014). Računalnolingvistička analiza korisničkih komentara na internetskim portalima. Diploma Thesis. Filozofski fakultet u Zagrebu, Department of Linguistics. [mentor Bekavac, Božo and Agić, Željko].

Preview

PDF (Croatian) Download (1MB) | Preview

Abstract

In September 2013 we collected users' comments on three Croatian news websites (jutarnji.hr, net.hr i bitno.net), made on articles about collecting signatures for a marriage referendum in Croatia. Comments were made between 12 May 2013 and 26 May 2013. We then conducted several computational linguistics analyses (sentiment analysis, polarity analysis and language analysis) on those comments. Results showed that comments on websites are mostly negative, with mostly negative sentiments and many negative polar words, phrases and sentences (the most positive website is Catholic website bitno.net). When it comes to quantitative analysis of sentiments and language, we found that there are statistically significant differences between general sentiment in comments and sentiment towards the initiative (the more positive sentiment towards the initiative, the more positive genral sentiment) and between general sentiment and language used in comments (the more positive general sentiment, the more standard the language). When it comes to language analysis, we found that around 50% of all comments was written in non-standard language variety, with many vulgarisms.

Actions (login required)

View Item