Kirsten Šteflovič

Kirsten Šteflovič

Auto-generated profile, based on publicly available information

Kirsten Šteflovič

University of Hradec Kralove

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 February 2024 in IEEE Access
Verified
Journal Article
Open access
JKJozef Kapusta
Jozef Kapusta
DDDávid Držík
Dávid Držík
KŠKirsten Šteflovič
Kirsten Šteflovič
KNKitti Szabó Nagy
Kitti Szabó Nagy
Contemporary language models heavily rely on large corpora for their training. The larger the corpus, the better a model can capture various semantic relationships. The issue at hand appears to be the limited scope of the corpora used. One potential solution to this problem is the application of data augmentation techniques to expand the existing corpus. Data augmentation encompasses several techniques for corpus augmentation. In this article, we delve deeper into the analysis of three techniques: Synonym Replacement, Back Translation, and Reduction of Function Words. Utilizing these three techniques, we prepared diverse versions of the corpus employed for training Word2Vec Skip-gram models. These techniques were validated through extrinsic evaluation, wherein Word2Vec Skip-gram models were used to generate word vectors for classifying fake news articles. Performance measures of the generated classifiers were analyzed. The study highlights significant statistically differences in classifier outcomes between augmented and original corpora. Specifically, Back Translation significantly enhances accuracy, notably with Support Vector and Bernoulli Naive Bayes models. Conversely, Reduction of Function Words (FWD) improves Logistic Regression, while the original corpus excels in Random Forest classification. The article also includes an intrinsic evaluation involving lexical semantic relations between word pairs. The intrinsic evaluation highlights nuanced differences in semantic relations across augmented corpora. Notably, the Back Translation (BT) corpus better aligns with established lexical resources, showcasing promising improvements in understanding specific semantic relationships.

1 June 2023 in Journal of Linguistics/jazykovedný Casopis
Verified
Journal Article
Open access
PDF file
DDDávid Držík
Dávid Držík
KŠKirsten Šteflovič
Kirsten Šteflovič
The utilization of text vectorization techniques has become essential for numerous classification tasks in present-day natural language processing. Word embedding methods commonly used today, such as Word2Vec, GloVe, etc., are based on the semantic similarity of words. WordNet, as a lexical database of words, provides a rich source of semantic information. In our article, we propose a text vectorization technique using extended text data with the data augmentation method, specifically by replacing words with their synonyms obtained from WordNet. The results obtained from text classification tasks using multiple classifiers demonstrate that expanding the corpus with this method leads to improved vector representations of words.

12 August 2025
Verified
Book Chapter
Closed access
PDF file
JHJan Hladěna
Jan Hladěna
KŠKirsten Šteflovič
Kirsten Šteflovič
PČPavel Čech
Pavel Čech
KŠKamila Štekerová
Kamila Štekerová
AŽAndrea Žváčková
Andrea Žváčková

15 August 2023 in Applied Sciences
Verified
Journal Article
Open access
PDF file
KŠKirsten Šteflovič
Kirsten Šteflovič
JKJozef Kapusta
Jozef Kapusta
There are several possibilities to improve classification in natural language processing tasks. In this article, we focused on the issue of coreference resolution that was applied to a manually annotated dataset of true and fake news. This dataset was used for the classification task of fake news detection. The research aimed to determine whether performing coreference resolution on the input data before classification or classifying them without performing coreference resolution is more effective. We also wanted to verify whether it is possible to enhance classifier performance metrics by incorporating coreference resolution into the data preparation process. A methodology was proposed, in which we described the implementation methods in detail, starting from the identification of entity mentions in the text using the neuralcoref algorithm, then through word-embedding models (TF–IDF, Doc2Vec), and finally to several machine learning methods. The result was a comparison of the implemented classifiers based on the performance metrics described in the theoretical part. The best result for accuracy was observed for the dataset with coreference resolution applied, which had a median value of 0.8149, while for the F1 score, the best result had a median value of 0.8101. However, the more important finding is that the processed data with the application of coreference resolution led to an improvement in performance metrics in the classification tasks.