This work addresses the task of text simplification for Kazakh, a morphologically rich and low-resource language. We propose KazSim, a fine-tuned simplification model based on large language models (LLM), using both the multilingual Llama series and Qwen2 backbones To support model training, we construct a parallel simplification dataset using a proposed Kazakh sentence complexity identification pipeline, selecting complex sentences from raw corpora with the proposed heuristic approach. As baselines, we include standard Seq2Seq models, Kazakh domain-specific large language models, and general-purpose instruction-following models in a zero-shot setup. Evaluation is performed on both an automatically constructed test set and a semi-manually created benchmark, using SARI, BLEU, ROUGE, and Bert-score. Results show that KazSim consistently outperforms all baselines, including domain-specific LLMs and zero-shot models, achieving good simplification quality while preserving meaning and controlling output length. We also examine the impact of prompt language on generation quality, comparing English and Kazakh instructions. While performance remains consistent overall, models tend to produce slightly better outputs when prompted in Kazakh, particularly in zero-shot and domain-specific settings.
This article presents an approach for spellchecking and autocorrection using web data for morphologically complex languages (in the case of Kazakh language), which can be considered an end-to-end approach that does not require any manually annotated word–error pairs. A sizable web of noisy data is crawled and used as a base to infer the knowledge of misspellings with their correct forms. Using the extracted corpus, a sub-string error model with a context model for morphologically complex languages are trained separately, then these two models are integrated with a regularization parameter. A sub-string alignment model is applied to extract symmetric and non-symmetric patterns in two sequences of word–error pairs. The model calculates the probability for symmetric and non-symmetric patterns of a given misspelling and its candidates to obtain a suggestion list. Based on the proposed method, a Kazakh Spellchecking and Autocorrection system is developed, which we refer to as QazSpell. Several experiments are conducted to evaluate the proposed approach from different angles. The results show that the proposed approach achieves a good outcome when only using the error model, and the performance is boosted after integrating the context model. In addition, the developed system, QazSpell, outperforms the commercial analogs in terms of overall accuracy.
The article describes the methods of identifying the conceptual content structure of the dataset of documents for the clustering. It was found that in the automatic extraction of key text concepts it is necessary to use the criteria of semantic significance of words and phrases obtained on the basis of syntactic, statistical and semantic methods. The syntactic criteria are based on the definition of the syntactic role of words and phrases in the text dataset. We accent on those elements of sentences that forms its semantic (predicate-actant) structure. In this research four methods of automatic identification of key text concepts have been elaborated, their comparative analysis is carried out and the technology of automatic creation of declarative means for text clustering of media is developed. The precision assessment of document clustering with and without declarative methods is conducted on test dataset.
The documents similarity metric is a substantial tool applied in areas such as determining topic in relation to documents, plagiarism detection, or problems necessary to capture the semantic, syntactic, or structural similarity of texts. Evaluated results of the similarity measure depend on the types of word represented and the problem statement and can be time-consuming. In this paper, we present a problem-independent algorithm of the similarity metric greedy texts similarity mapping (GTSM), which is computationally efficient to be applied for large datasets with any preferred word vectorization models. GTSM maps words in two texts based on a decision rule that evaluates word similarity and their importance to the texts. We compare it with the well-known word mover’s distance (WMD) algorithm in the k-nearest neighbors text classification problem and find that it leads to similar or better results. In the correlation evaluation task of similarity measures with human-judged scores, we demonstrate its higher correlation scores in comparison with WMD and sentence mover’s similarity (SMS) and show that GTSM is a decent alternative for both word-level and sentence-level tasks.
This paper addresses text simplification task for Kazakh, a morphologically rich, low-resource language, by introducing KazSim, an instruction-tuned model built on multilingual large language models (LLMs). First, we develop a heuristic pipeline to identify complex Kazakh sentences, manually validating its performance on 400 examples and comparing it against a purely LLM-based selection method; we then use this pipeline to assemble a parallel corpus of 8709 complex–simple pairs via LLM augmentation. For the simplification task, we benchmark KazSim against standard Seq2Seq systems, domain-adapted Kazakh LLMs, and zero-shot instruction-following models. On an automatically constructed test set, KazSim (Llama-3.3-70B) achieves BLEU 33.50, SARI 56.38, and F1 87.56 with a length ratio of 0.98, outperforming all baselines. We also explore prompt language (English vs. Kazakh) and conduct human evaluation with three native speakers: KazSim scores 4.08 for fluency, 4.09 for meaning preservation, and 4.42 for simplicity—significantly above GPT-4o-mini. Error analysis shows that remaining failures cluster into tone change, tense change, and semantic drift, reflecting Kazakh’s agglutinative morphology and flexible syntax.
30 September 2018 in Международный научный журнал "Современные информационные технологии и ИТ-образование"
We consider generalization as a property of human thinking to make general conclusion based on authors’ own experience and observations and one of the techniques of authors use to manipulate the readership and present an algorithm for evaluation of the generalization in texts. The algorithm is based on the lexicon-based approach. To search the generalization we use ready-made dictionary (KEY-dictionary) and RuSentiLex dictionary. KEY-dictionary contains words and phrases (elements) that express the generalization. In RuSentiLex we take the words and phrases that express opinion and fact. The algorithm searches exact matches the elements from text with the elements from the dictionaries, it is also important that the elements from different dictionaries have their weights. New method is developed for automatic detection of generalization in texts from official media. Numerical calculations of generalization were performed using a special software application. To test the proposed approach the expert estimation were used. Статья представляет обобщение как свойство человеческого мышления делать общий вывод на основе собственного опыта и наблюдений авторов, также рассматривается один из приемов, используемых авторами для манипулирования читательской аудиторией, и представлен алгоритм оценки обобщения в текстах. Алгоритм основан на лексиконном подходе. Для поиска обобщения используется готовый словарь (KEY-dictionary)и словарь RuSentiLex. KEY -словарь содержит слова и словосочетания (элементы), выражающие обобщение. В RuSentiLex мы берем слова и фразы, которые выражают мнение и факт. Алгоритм поиска точно сопоставляет элементы из текста с элементами из словарей, также важно, чтобы элементы из разных словарей имели свои веса. Разработан новый метод автоматического обнаружения обобщения в текстах официальных СМИ. Численные расчеты обобщения выполнены с использованием специального программного приложения. Для проверки предложенного подхода использовалась экспертная оценка.