Intelligent Decision Support Systems for Sustainable Computing

Volume 705 of the series Studies in Computational Intelligence pp 207-224


Analyzing Slavic Textual Sentiment Using Deep Convolutional Neural Networks

  • Leo MršićAffiliated withUniversity College Algebra Email author 
  • , Robert KopalAffiliated withUniversity College Algebra
  • , Goran KlepacAffiliated withUniversity College Algebra

* Final gross prices may vary according to local VAT.

Get Access


In this paper we deploy a deep architecture for convolutional neural networks for understanding of Croatian. We follow the same approach as in [34], and we use deep learning on character inputs on a sentiment analysis dataset in Croatian. Although we have archived considerable results (without any complex parsing or background knowledge), the result was inferior to that reported in the abovementioned paper. As Croatian is one of the low-resource languages, there are considerable links between using such an approach (that maximizes the role of data) and sustainability. The main objective of this chapter is to give a clear understanding of the position of low-resource languages and propose a direction for sustainable development of language technologies illustrated using convolutional neural networks for textual sentiment analysis . Impact of this research to scientific but also business community is significant due to fact that every method with acceptable ratio of simplicity and effectiveness can be included inside more complex logic environment focusing on specific language or appliance in specific area. Since language structure is generally not easy to manage, there is constant need for improvement of tools used to score text data especially while majority of unstructured data analysis tools often transfer various data to text and create further analysis paths form there.


Deep convolutional neural networks Sentiment analysis Slavic languages Sustainable techniques