View on GitHub


A Sentiment Analysis Dataset of Comments in Serbian

SentiComments.SR - A Sentiment Analysis Dataset of Comments in Serbian

The SentiComments.SR dataset includes the following three corpora:

Dataset creation

The main SentiComments.SR corpus was constructed out of the comments written by visitors on the movie review website in Serbian. Each comment was assigned a unique ID based on the movie to which it referred, and its placement in the comment tree for that particular movie, signifying its position within the whole discussion. Comments containing more than a predefined upper bound for token count (using basic whitespace tokenization), were discarded, as were the comments not written in Serbian.

The movie verification corpus comments were sourced from two other Serbian movie review websites - and The book verification corpus comments were also sourced from the website.

Dataset annotation

Six sentiment labels were used in dataset annotation: +1, -1, +M, -M, +NS, and -NS, with the addition of an ‘s’ label suffix denoting the presence of sarcasm. The annotation principles used to assign sentiment labels to items in SentiComments.SR are described in the papers listed in the References section.

The main SentiComments.SR corpus was annotated by two annotators working together, and therefore contains a single, unified sentiment label for each comment. The verification corpora were used to evaluate the quality, efficiency, and cost-effectiveness of the annotation framework, which is why they contain separate sentiment labels for six annotators, divided into three groups - initial group (#1 and #2), experimental group (#3 and #4) and control group (#5 and #6).

Dataset format

The main SentiComments.SR corpus is available as a tab-separated .txt file, in two variants:

Both variants have the following column structure, and they both share the same sentiment labels and comment IDs:

The verification corpora are also available as tab-separated .txt files:

The column structure of the verification corpora files is as follows:

Comments in all files are written in either the Serbian Latin or the Serbian Cyrillic script. All files are encoded in UTF-8.


If you wish to use the SentiComments.SR dataset (or the annotation principles applied in its construction) in your paper or project, please cite the following paper:

The reference for the full annotation guidelines (in Serbian) is the following:


The annotation of the SentiComments.SR dataset was supported by the Regional Linguistic Data Initiative (ReLDI) via the Swiss National Science Foundation grant no. 160501.


The SentiComments.SR dataset is available under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license. The entire text of the license can be found here.

If you wish to use this dataset for a commercial product, please contact me at: vuk.batanovic / at /