ResearchClass: Mining Sentiment Classification from Slovenian News Texts

Datum održavanja: petak, 5.2.2016. u 12:15 h u prostoriji O-357
Predavač: doktorand Jože Bučar (Faculty of Information Studies, Novo mesto)
Naziv predavanja: Mining Sentiment Classification from Slovenian News Texts


When the World Wide Web came along in 1990, it has radically changed lifestyles and functioning of individuals, organizations, and society. The exponential growth of data and users has lead the Web to become the main medium and an impressive source of data for every data scientist. Nowadays the Web is a huge  repository of structured and unstructured data; it contains facts and opinions, which we can extract.

In this talk we will describe a background of retrieving textual documents with political, business, economic, and financial content from five Slovenian web media between the September 2007 and December 2013. We generated stratified random sample of 10,427 documents.  Annotation of documents, which were labelled into three different categories (negative, neutral and positive), was carried out on three  levels independently, i.e. document-level, paragraph-level and sentence-level. To evaluate the process of annotation, we will present the inter-annotator agreement. We will also provide some examples of its use, evaluation performance of sentiment based classification techniques, estimation of the proportion of negative, neutral, and positive news in web media, and finally, monitoring the topic-sentiment dynamics.


Jože Bučar is a Young Researcher at the Laboratory of Data Technologies at the Faculty of Information Studies. His primary research interests are in applications of data mining (text and web mining, sentiment analysis) as data analysis, decision support, machine learning  and statistical analysis of complex networks. Jože Bučar studied geodesy at the Faculty of Civil and Geodetic Engineering at University of  Ljubljana, Slovenia, where he received M.Sc. in 2011. He is currently a PhD student at the Faculty of information studies in Novo mesto.