This study focuses on the extraction and semantic analysis of data from tables, emphasizing the importance of understanding the semantics of tables to obtain useful information. The main goal was to develop a technology using the ontology for the semantic analysis of tables. An iterative algorithm has been proposed that can parse the contents of a table and determine cell types based on the ontology. The study presents an automated method for extracting data in various languages in various fields, subject to the availability of an appropriate ontology. Advanced techniques such as cosine distance search and table subject classification based on a neural network have been integrated to increase efficiency. The result is a software application capable of semantically classifying tabular data, facilitating the rapid transition of information from tables to ontologies. Rigorous testing, including 30 tables in the field of water resources and socio-economic indicators of Kazakhstan, confirmed the reliability of the algorithm. The results demonstrate high accuracy with a notable triple extraction recall of 99.4%. The use of Levenshtein distance for matching entities and ontology as a source of information was key to achieving these metrics. The study offers a promising tool for efficiently extracting data from tables.

In this paper, we propose a pipeline aimed at automatically extracting tables from heterogeneousWeb sources, such as HTML pages, pdf files and images. Table extraction is one of the activelydeveloping areas of Information Extraction, for which many applications, libraries and frameworksare currently being developed. Nevertheless, most of these tools are focused on solving somespecific tasks, for example, only on recognizing tables presented in the form of images. Wepropose combining these tasks into a single pipeline that will support the full cycle of processingtables – starting with the stages of their search, recognition and extraction and ending with thestages of semantic analysis and interpretation, i.e. understanding tables (table understanding).Understanding tables and replenishing (population) knowledge bases (knowledge graphs) withmeaningful information contained in these tables is the ultimate goal of our design. The firstpart of the work presents methods for detecting tables on web pages, pdf documents, as well asautomatic detection of attributes and values of objects. The second part presents the architectureof the Qurma tool and its structure. The results show the implementation of the parser for theAlmaty-Ust-Kamenogorsk air search theme.

In this study, we developed a model for automatic recognition of Kazakh speech by fine-tuning the XLSR-Wav2Vec2 pre-trained model to a corpus of Kazakh speech. Our results show that fine-tuning the wav2vec2 model on a small corpus of Kazakh speech allows a significant increase in recognition accuracy. However, larger datasets are needed to further evaluate the effectiveness of this approach. The results of this study contribute to ongoing efforts to improve speech recognition technology for low-resource languages such as Kazakh.

This study presents a rule-based hybrid pipeline for the automated extraction of structured metadata from PDF versions of Kazakh-language newspaper articles, focusing on the national newspaper Egemen Qazaqstan. The primary goal is to support the development of a machine-readable knowledge base for future use in training large language models (LLMs) and building an AI-powered assistant for data journalism in Kazakhstan. The pipeline integrates three open-source parsers – pdfminer.six, PyMuPDF, and pdfplumber – to extract key elements such as title, author, date, abstract, text, journal name, and category. To evaluate extraction quality, we compared the results of the automated parser against manually annotated reference files across three real-world issues of the newspaper. The evaluation employed three complementary metrics: Precision, Textual Semantic Similarity (TSS), and Holistic Precision, which jointly assess both exact and semantic matches. The experimental results show that most tags – especially structured fields like date, journal, and category – achieved perfect Holistic Precision (1.00), while more variable fields like title still scored above 0.85. The validated pipeline was then applied to the full corpus of 2,140 newspaper PDFs published between 2017 and March 2025, successfully converting 159,135 articles into structured JSON format. This enriched corpus serves as a foundational knowledge base for Kazakh-language AI systems in journalism and media analysis.