A Comparative Analysis of LSTM and BERT Models for Named Entity Recognition in Kazakh Language: A Multi-classification Approach

This study introduces a new method for solving multi-classification tasks in the Kazakh language using LSTM and BERT models, based on datasets for named entity recognition (NER). It addresses the critical issue of limited data for the Kazakh language by adapting existing resources to explore the details of text classification. To address this gap, our study leverages the potential of existing pretrained models such as BERT, adapting methodologies from well-resourced languages. This approach utilizes the linguistic richness of Kazakh by employing sophisticated NLP strategies, enabling effective text classification despite limited available datasets. The research shows the strong performance of LSTM and BERT models in handling multi-dimensional classification tasks, making a significant contribution to the progress of natural language processing (NLP) for languages with limited datasets. The methodology in this study highlights the flexibility of modern NLP models and opens new possibilities for practical use. By using adapted datasets, the research demonstrates how these models can solve complex language problems, ensuring more accurate and detailed text classification. This approach shows the potential of advanced NLP techniques to improve the processing and understanding of underrepresented languages. Additionally, the study provides a thorough evaluation of the performance of LSTM and BERT models in the context of Kazakh, offering insights into their strengths and areas for improvement. By integrating these models into the NER framework, the research creates more robust and versatile applications, enhancing the overall effectiveness of NLP technologies in multilingual and resource-limited environments. This broader focus not only contributes to the theoretical understanding of NLP but also has practical implications for developing more inclusive and effective language processing tools.