HYBRID METHOD FOR INTELLIGENT TEXT CONTENT VERIFICATION BASED ON LEXICON ANALYSIS AND THE BERT MODEL
Infocommunication technologies and electronic engineering · 2026-11 · 6 (2) · 63-75
DOI: 10.23939/ictee2026.02.063 ↗Kurz gefasst
Fachartikel in der Zeitschrift Infocommunication technologies and electronic engineering, 2026-10. Verfasst von Mykola Beshley und 1 weiteren Autor:innen. Die Arbeit nennt 18 Literaturangaben.
Abstract (Originalsprache)
The paper proposes a hybrid method for intelligent text content verification aimed at the automated detection of toxicity, psychological triggers, emotionally sensitive content, and potentially harmful textual information. The relevance of the study is determined by the continuous growth of textual information in the digital environment and the need for rapid, context-aware analysis, since the presence of a single undesirable lexical unit does not always indicate that the entire message contains harmful or sensitive content. The purpose of the research is to develop a method and software system for intelligent text analysis that combines fast, lexicon-based detection of explicit markers with contextual classification using the BERT transformer model, and provides personalised verification results. To achieve this purpose, modern approaches to natural language processing, automatic toxicity detection, and the identification of psychologically sensitive content were analysed, and the advantages and limitations of lexicon-based and neural-network methods were considered. A two-stage analysis procedure is proposed. In the first stage, the input text is processed using a lexicon-based approach to rapidly identify explicit lexical markers associated with predefined categories of potentially sensitive content. At the second stage, contextual analysis is performed using the BERT model, enabling consideration of the semantic context of words and phrases and the identification of characteristics that cannot be reliably detected through direct lexical matching. The outputs of both stages are aggregated into generalised indicators and transformed into an interpretable verification report. Based on the proposed approach, the BetweenLines software system was developed using a client-server architecture. The client application was implemented with Flutter, while the server component was developed using the FastAPI framework and provides access to the intelligent text analysis module. The embedded relational database SQLite is used for authentication, storage of analysis results, verification history, and personalised user settings. An important feature of the system is the ability to configure active content categories and sensitivity thresholds, enabling the verification process to be tailored to individual user requirements. The system also provides structured reporting and access to previously performed analyses, thereby supporting repeated verification and comparison of results. Experimental testing was conducted in both mobile and web environments to evaluate interactions among the main software components, request processing, result storage, error handling, and personalisation mechanisms. The obtained results confirmed the operability of the proposed approach and demonstrated the practical feasibility of combining lexicon-based and contextual neural analysis for intelligent text content verification. During testing, the average processing time was 1–3 seconds, depending on the amount of input text and server workload, while the achieved classification accuracy exceeded 88%. The proposed method and software architecture can serve as a basis for intelligent digital content analysis tools, personalised filtering systems, and user-support applications designed to improve awareness and safety in interacting with potentially sensitive textual information.
Zitieren
APA
Beshley, M., & Chyrun, S. (2026). HYBRID METHOD FOR INTELLIGENT TEXT CONTENT VERIFICATION BASED ON LEXICON ANALYSIS AND THE BERT MODEL. Infocommunication technologies and electronic engineering, 6(2), 63-75. https://doi.org/10.23939/ictee2026.02.063
Vancouver
Beshley M, Chyrun S. HYBRID METHOD FOR INTELLIGENT TEXT CONTENT VERIFICATION BASED ON LEXICON ANALYSIS AND THE BERT MODEL. Infocommunication technologies and electronic engineering. 2026;6(2):63-75. doi:10.23939/ictee2026.02.063
BibTeX
@article{Beshley2026,
author = {Beshley, Mykola and Chyrun, Sofia},
title = {HYBRID METHOD FOR INTELLIGENT TEXT CONTENT VERIFICATION BASED ON LEXICON ANALYSIS AND THE BERT MODEL},
journal = {Infocommunication technologies and electronic engineering},
year = {2026},
volume = {6},
number = {2},
pages = {63--75},
doi = {10.23939/ictee2026.02.063}
}
EUVIMED