Papers from conferences and scientific journals will be published in this section.

• V.Batanović et al., "Open Resources and Technologies for Serbian Language Processing"

  V.Batanović, N.Ljubešić, T. Samardžić, M. Miličević Petrović, "Open Resources and Technologies for Serbian Language Processing" (in Serbian), PSSOH conference, Belgrade, Oct. 2020
  Link: https://zenodo.org/record/4113230#.X6GcaohKiUk
  Abstract: The openness of language resources and tools is of great importance for increasing the quality and speed of development of technologies for computer processing of natural languages. This paper presents open resources for the processing of the Serbian language. Hand-annotated corpora are described, as well as a wider range of tools and computer models, including a web service that makes them easy to use.

• V.Batanović et al., "A versatile framework for resource-limited sentiment articulation, annotation, and analysis of short texts"

  V. Batanović, M.Cvetanović, B.Nikolić, "A versatile framework for resource-limited sentiment articulation, annotation, and analysis of short texts", PLoS ONE 15(11): e0242050. https://doi.org/10.1371/journal.pone.0242050
  Link: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0242050
  Abstract:
  Choosing a comprehensive and cost-effective way of articulating and annotating the sentiment of a text is not a trivial task, particularly when dealing with short texts, in which sentiment can be expressed through a wide variety of linguistic and rhetorical phenomena. This problem is especially conspicuous in resource-limited settings and languages, where design options are restricted either in terms of manpower and financial means required to produce appropriate sentiment analysis resources, or in terms of available language tools, or both. In this paper, we present a versatile approach to addressing this issue, based on multiple interpretations of sentiment labels that encode information regarding the polarity, subjectivity, and ambiguity of a text, as well as the presence of sarcasm or a mixture of sentiments. We demonstrate its use on Serbian, a resource-limited language, via the creation of a main sentiment analysis dataset focused on movie comments, and two smaller datasets belonging to the movie and book domains. In addition to measuring the quality of the annotation process, we propose a novel metric to validate its cost-effectiveness. Finally, the practicality of our approach is further validated by training, evaluating, and determining the optimal configurations of several different kinds of machine-learning models on a range of sentiment classification tasks using the produced dataset.

• S.Tubić et al., "Annotated functional decomposition"

  S. Tubić, M. Cvetanović, Z. Radivojević, S. Stojanović, "Annotated functional decomposition", COMPUTER APPLICATIONS IN ENGINEERING EDUCATION, pp. 1-13, March, 2021
  Link: https://onlinelibrary.wiley.com/doi/10.1002/cae.22394
  Abstract: Experiences gained from the domain‐specific courses showed that students focus mostly on how to implement solutions and less on what must be considered within the solution. In the case of information systems-related courses, students focus on system development using specific languages and frameworks while often disregard the required logical checks and constraints. This paper introduces annotated functional decomposition (AFD) to assist students in overcoming the challenge of understanding the logic of an information system. AFD leverages methodological concepts from computational thinking and represents a problem decomposition approach that is extended with additional levels of decomposition. These levels of decomposition are orthogonal and implemented with annotations that enrich a decomposition with information regarding control and data flow, as well as reuse and implementation details. AFD could be exercised with a supporting AFD Tool developed as an Eclipse IDE plugin that performs syntax and semantic checks along with the generation of UML sequential diagrams. The AFD Tool and its source code are available free of charge. Quantitative and qualitative evaluations of AFD Tool usage during an information systems course revealed that students who used AFD achieved higher average grades than those who used UML for solving the same problems, and moreover that students perceived AFD as easy to understand and use.

• M.Kotlar et al., "Novel Meta-Features for Automated Machine Learning Model Selection in Anomaly Detection"

  M. Kotlar, M. Punt, Z. Radivojević, M. Cvetanović, V.Milutinović, "Novel Meta-Features for Automated Machine Learning Model Selection in Anomaly Detection", IEEE ACCESS (Volume: 3), pp. 89675 - 89687, June, 2021
  Link: https://ieeexplore.ieee.org/document/9461173
  Abstract: A growing number of research papers shed light on automated machine learning (AutoML) frameworks, which are becoming a promising solution for building complex machine learning models without human expertise and assistance. The key challenge in enabling AutoML frameworks to build an efficient model for anomaly detection tasks is to determine the best underlying model for a given task and optimization metric. The meta-learning approaches based on a set of meta-features that describes data properties can enable efficient model selection in AutoML frameworks. The existing meta-learning approaches based on statistical and information-theoretic meta-features require large amounts of data and computational resources to extract data properties. This paper proposes a novel set of meta-features for model selection in anomaly detection tasks based on domain-specific properties of data which overcomes the shortcomings of existing meta-features by introducing simple but effective meta-features that can be efficiently extracted or estimated by using a low amount of data. Experiments with 63 datasets from different repositories with varying schemas show that the proposed set of meta-features achieves an accuracy of 87% for model selection, while the achieved accuracy for simple meta-features is 74%, for statistical meta-features 68%, for information theory meta-feature 70%, and for a comprehensive set of meta-features by pyMFE 73%. This demonstrates that the proposed set can be adopted by AutoML frameworks across a diverse range of domains.

• V. Batanović, "Semantic Similarity and Sentiment Analysis of Short Texts in Serbian"

  V. Batanović, "Semantic Similarity and Sentiment Analysis of Short Texts in Serbian", 29^th Telecommunications Forum "TELFOR 2021", IEEE Serbia & Montenegro, November 2021
  Link: https://www.telfor.rs/program-naucne-sekcije
  Abstract: This paper presents an overview of the open access datasets in Serbian that have been manually annotated for the tasks of semantic textual similarity and short-text sentiment classification. In addition, it describes several kinds of statistical models that have been trained and evaluated on these datasets and discusses their results.

• V.Otašević et al., "Evaluation of text messages using convolutional neural networks"

  V.Otašević, D.Drašković, B.Nikolić, "Evaluation of text messages using convolutional neural networks", 29^th Telecommunications Forum "TELFOR 2021", IEEE Serbia & Montenegro, November 2021
  Link: https://www.telfor.rs/program-naucne-sekcije
  Abstract: The impossibility of defining the numerical value of text is a major issue when it is necessary to analyze users’ comments and feedbacks about provided services. This paper presents a tool able to overcome this issue. The developed tool, which evaluates text messages, is based on a convolutional neural network. The main goal of this paper is to present the result of the research and tool’s performance in solving the problem of determining the value of text messages.

• I.Munjas, V.Batanović, "US address classification based on text processing and machine learning"

  I. Munjas, V.Batanović, "US address classification based on text processing and machine learning", 29^th Telecommunications Forum "TELFOR 2021", IEEE Serbia & Montenegro, November 2021
  Link: https://www.telfor.rs/program-naucne-sekcije
  Abstract: Addresses represent a crucial type of textual data for real estate companies. In order to identify, fix, or remove incorrect entries, we categorize addresses into one of six predefined classes. In this context, we explore the effects of different text processing and classification methods. The best results are obtained by using non-linear classifiers with a combination of unigram and bigram features.

• T.Šekularac, D.Drašković, "Software system for improving communication of children with disabilities in the Serbian language"

  T.Šekularac, D.Drašković, "Software system for improving communication of children with disabilities in the Serbian language", 29^th Telecommunications Forum "TELFOR 2021", IEEE Serbia & Montenegro, November 2021
  Link: https://www.telfor.rs/program-naucne-sekcije
  Abstract: This paper presents a software system designed as communication aid for children with disabilities. The system generates digitized speech using a dictionary of symbols in the matrix form, which can be extended and modified, as well as the corresponding audio recordings and visual representations of symbols. The creators of this system managed to eliminate the shortcomings of the existing software solutions, namely the lack of flexibility and accessibility, and inability to cater for individual differences. The system is realized as a multiplatform application using Xamarin technology.