Title: Sentiment annotated news corpus and sentiment lexicon in Slovene
Author: Jože Bučar, Faculty of Information Studies Novo mesto (contact: joze.bucar@gmail.com)
Abstract: This work deals with a construction of Manually sentiment annotated Slovenian news corpus (SentiNews 1.0), Automatically sentiment annotated Slovenian news corpus (AutoSentiNews 1.0) and Slovene sentiment lexicon (JOB 1.0). It also provides the tools that were required for their construction. Web crawlers, written in R language, retrieve Slovenian news texts from the portals 24ur, Dnevnik, Finance, Rtvslo, and Žurnal24. These portals contain political, business, economic and financial content. The AutoSentiNews 1.0 contains 256,567 documents that were published between 1 September 2007 and 31 January 2016. Between 2 and 6 annotators independently annotated 10,427 documents (SentiNews 1.0), which were published between 1 September 2007 and 31 December 2013, as negative, neutral and positive on three levels, i.e. document, paragraph, and sentence level using the five-level Lickert scale (1 – very negative, 2 – negative, 3 – neutral, 4 – positive, and 5 – very positive). The remaining 246,140 news were annotated automatically. The lexicon (JOB 1.0) contains a list of 25,524 headwords from the List of Slovenian headwords 1.1 (http://hdl.handle.net/11356/1038) extended with sentiment ratings based on the AFINN model with an integer between -5 (very negative) and +5 (very positive). The ratings are derived from the lemmatized version of the (sentence-based) SentiNews 1.0 by deducting the average sentiment of the corpus.
Keywords: Web-crawling, R, News corpus, Sentiment classification, Opinion mining, Sentiment lexicon, Opinion lexicon
Web resources:
- Slovenian news texts with political, business, economic and financial content published between 1 September 2007 and 31 January 2016 from five Slovenian web media from five web media: www.24ur.com, www.dnevnik.si, www.finance.si, www.rtvslo.si, www.zurnal24.si
- ToTaLe text analyser for Slovene texts - Tokanizer, Tagger and Lemmatizer for Slovene texts, Department of Knowledge Technologies, Jožef Stefan Institute (http://nl.ijs.si/analyse/)
- List of Slovenian words, Fran Ramovš Institute of Slovenian Language ZRC SAZU (http://bos.zrc-sazu.si/besede_en.html)
Type: .txt, .R, .zip
Encoding: UTF-8
Year: 2017-05-09
Acknowledgements: The authors gratefully acknowledge Fran Ramovš Institute of Slovenian Language ZRC SAZU for the web list of Slovenian words and Department of Knowledge Technologies, Jožef Stefan Institute for the ToTaLe text analyser for Slovene texts. We also thank Dejan Zidar for his assistance at building the web application and annotators for outstanding work in the context of the annotation process. This paper is based upon work supported by the European Union, The European Regional Development Fund, Slovene Human Resources Development and Scholarship Fund, Ministry of Education, Science and Sport, Slovenia and Young Researcher Programme by Slovenian Research Agency. Our research was carried out within the framework of the Operational Programme for Strengthening Regional Development Potentials for the period 2007-2013, Development Priority 1: Competitiveness and research excellence, Priority Guideline 1.1: Improving the competitive skills and research excellence.
Citation: If you are a scientist and use the wordlists or code you can cite this publication:
@article{buvcar2018annotated, title={Annotated news corpora and a lexicon for sentiment analysis in Slovene}, author={Bu{\v{c}}ar, Jo{\v{z}}e and {\v{Z}}nidar{\v{s}}i{\v{c}}, Martin and Povh, Janez}, journal={Language Resources and Evaluation}, volume={52}, number={3}, pages={895--919}, year={2018}, publisher={Springer} }
or
@inproceedings{buvcar2016sentiment, title={Sentiment Classification of the Slovenian News Texts}, author={Bu{\v{c}}ar, Jo{\v{z}}e and Povh, Janez and {\v{Z}}nidar{\v{s}}i{\v{c}}, Martin}, booktitle={Proceedings of the 9th International Conference on Computer Recognition Systems CORES 2015}, pages={777--787}, year={2016}, organization={Springer} }
Copyright: Web media are credited for publishing the news. Web texts, that we required within our project, are under their copyrights, therefore they are available for non-commercial scientific and research purposes only. Research can be performed by a faculty member at an accredited academic institution (college or university) or a research institute member; wherein the research (i) does not now or in the future benefit, or does not now or in the future involve, or is not funded by, a commercial entity; and/or (ii) is not now, or in the future, subject to consulting or licensing obligations or other grant of rights to any commercial entity or third party, and (iii) will not generate any intellectual property rights for any one or more of the faculty member, academic institution, third party, or commercial entity, and (iv) the results of which will be released in the public domain by publication. If your research does not meet these criteria, please contact the author.
Through our work we support the open source community, in order to allow future researchers to contribute to (computational) linguistics community. Thus, some of our work, such as web crawlers, manual and automatic annotations of the news and a lexicon for sentiment analysis in Slovene, is publicly available under Creative Commons copyright license Attribution-ShareAlike 4.0 International (CC BY-SA 4.0 or newer version). This license allows others to use, distribute, reproduce, and build upon this work as long as they credit this work by clearly mentioning its author and title. Moreover, they can license their new work under the identical terms, so any derivatives will also allow commercial use. For more details, check out the information available on the website https://creativecommons.org/licenses/by-sa/4.0/.