Journal "Software Engineering"
a journal on theoretical and applied science and technology
ISSN 2220-3397

Issue N5 2025 year

DOI: 10.17587/prin.16.260-268
Implementation of a Level of Access to a Language Model Using the Example of an Intelligent Applicant Support System
E. A. Kaledina, Associate Professor, elena.lizina@gmail.com, O. E. Kaledin, Associate Professor, kaledinoe@gmail.com, A. V. Artamonov, Master's Student, alexey375890@gmail.com, National Research Mordovian State University N. P. Ogareva, Saransk, 430005, Russian Federation
Corresponding author: Elena A. Kaledina, Associate Professor Department of Data Analysis and Artificial Intelligence, National Research Mordovian State University N. P. Ogareva, Saransk, 430005, Russian Federation E-mail: elena.lizina@gmail.com
Received on January 30, 2025
Accepted on March 18, 2025

The paper proposes models for segmenting text documents to implement fine-tuning of large language models using the RAG (Retrival-Augmented Generation) method. In the first model, the document is divided into a large number of fragments equal in number of characters. In this case, the partition is carried out in such a way that for any fragment there are either one or two fragments that coincide exactly by half. In the second model, the source document is divided into fragments according to the semantic content. In turn, the fragments are divided into segments, resulting in a two-level partition of the document. Based on the concept of two-level partitioning, a "Key-Value" structure is created, where the keys are text segments, and the value is the corresponding text fragment. Using the example of implementing an intelligent information system to support university applicants, a comparison and analysis of these models was carried out. The information system includes a database, web interface and chat bot. Giga chat from the Sber company was used as a large language model. Fine-tuning is carried out on the basis of data provided by the National Research Ogarev Mordovia State University. The work describes the presentation model and document structure for further work with the chatbot, as well as a vector storage that provides a convenient interface for working with vectorized data from the outside. The developed language model access layer provides an abstraction over the connected chatbot, while providing the user with a fairly low-level degree of interaction with the generative model, which can be useful when constructing complex queries.

Keywords: large language model, fine-tuning of the model, fragmentation of text documents, advanced search generation
pp. 260—268
For citation:
Kaledina E. A., Kaledin O. E., Artamonov A. V. Implementation of a Level of Access to a Language Model Using the Example of an Intelligent Applicant Support System, Programmnaya Ingeneria, 2025, vol. 16, no. 5, pp. 260—268. DOI: 10.17587/prin.16.260-268 (in Russian).
References:
  1. Johri P., Khatri S. K., Al-Taani A. T. et al. Natural Language Processing: History, Evolution, Application, and Future Work, Lecture Notes in Networks and Systems, 2021, vol. 167, pp. 365—375. DOI 10.1007/978-981-15-9712-1_31.
  2. Zhao W. X., Zhou K., Li J. et al. A Survey of Large Language Models, available at: https://arxiv.org/abs/2303.18223 (date of access 23.01.2025).
  3. Lehmann J., Meloni A., Motta E. et al. Large Language Models for Scientific Question Answering: An Extensive Analysis of the SciQA Benchmark, In book: The Semantic Web, 2024, pp. 199—217. DOI: 10.1007/978-3-031-60626-7_11.
  4. Liu E., Cui C., Zheng K., Neubig G. Testing the Ability of Language Models to Interpret Figurative Language, available at: https://arxiv.org/abs/2204.12632 (date of access 10.03.2025).
  5. Fortino A., Yang Z. Evaluating Large Language Model Accuracy in Structured Academic Settings: Three Case Studies, Conference: IEEE ISEC'24 14th IEEE Integrated STEM Education Conference. At: Princeton University, 2024. DOI: 10.1109/ISEC61299.2024.10665280.
  6. Galeev D. T., Panishchev V. S. Experimental study of "transformer" language models in the task of finding an answer to a question in a Russian-language text, Informatika i avtomatizatsiya, 2022, vol. 21, no. 3, pp. 521—542. DOI: 10.15622/ia.21.3.3 (in Russian).
  7. Gritsai G. M., Grabovoy A. V., Kildyakov A. S., Chekho-vich Yu. V. Search for artificially generated text fragments in scientific documents, Doklady Rossiyskoy akademii nauk. Matematika, informatika, protsessy upravleniya, 2023, vol. 514, no. 2, pp. 308—317. DOI: 10.31857/S2686954323601677 (in Russian).
  8. Ziyue V. Optimization and development potential of machine translation using the example of large language models such as Chat-GPT, Aktual'nyye voprosy perevodovedeniya i regionovedeniya: sbornik statey Mezhdunarodnogo foruma, Nizhniy Novgorod: Nizhegorodskiy gosudarstvennyy lingvisticheskiy universitet im. N. A. Dobrolyubova, 2024, pp. 67—75 (in Russian).
  9. Khusainova A. M., Romanov V. A., Khan A. M. Multilingual machine translation using a hierarchical transformer, Vestnik Voronezhskogo gosudarstvennogo universiteta. Seriya: Sistemnyy analiz i informatsionnyye tekhnologii, 2022, no. 1, pp. 125—138. DOI: 10.17308/sait.2022.1/9207 (in Russian).
  10. Vaswani A., Shazeer N., Parmar N. et al. Attention is all you need, Proceedings of the 31st International Conference on Neural Information Processing Systems, Curran Associates Inc., Red Hook, NY, USA, 2017, pp. 6000—6010.
  11. Han B., Susnjak T., Mathrani A. Automating Systematic Literature Reviews with Retrieval-Augmented Generation: A Comprehensive Overview, Applied Sciences, 2024, vol. 14, article 9103. DOI: 10.3390/app1419910313.
  12. Mekhdiyev E. T., Borisov A. A., Rostotskiy M. V. et al. Development of an intelligent chat bot to automate responses and analyze user needs: computer science and computer technology, Human Progress, 2024, vol. 10, no. 5 (in Russian).
  13. Bora A., Cuayahuitl H. Systematic Analysis of Retrieval-Augmented Generation-Based LLMs for Medical Chatbot Applications, Machine Learning and Knowledge Extraction, 2024, vol. 6, no. 4, pp. 2355—2374. DOI 10.3390/make6040116.
  14. Zhou D., Scharli N., Hou L. et al. Least-to-Most Prompting Enables Complex Reasoning in Large Language Models, available at: https://arxiv.org/abs/2205.10625 (date of access 23.01.2025).
  15. Devlin J., Chang M.-W., Lee K., Toutanova K. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, available at: https://arxiv.org/abs/1810.04805 (date of access 23.01.2025).
  16. Gao Y., Xiong Y., Gao X. et al. Retrieval-Augmented Gen­eration for Large Language Models: A Survey, available at: https://arxiv.org/abs/2312.10997 (date of access 23.01.2025).
  17. Nazarov D. M., Begicheva S. V. Application of large language models in the educational process, Biznes. Obrazovani-ye. Pravo, 2024, no 3(68), pp. 430—436. DOI: 10.25683/VOL-BI.2024.68.1057 (in Russian).
  18. Obolensky D. M., Shevchenko V. I. Using the RAG method and large language models in intellectual educational ecosystems, Ekonomika. Informatika. 2024, vol. 51, no. 3, pp. 699—709. DOI: 10.52575/2687-0932-2024-51-3-699-709 (in Russian).
  19. Abramov M. V., Khlobystova A. O., Sazanov V. A. Development of a chatbot for tracking the table of classes at the university, Programmnyye produkty i sistemy, 2023, vol. 36, no. 3, pp. 466—473. DOI: 10.15827/0236-235X.143.466-473 (in Russian).
  20. Safronova M. A., Starina A. A. Chat bot in telegram — an intelligent assistant in the digital educational environment of a university for foreign students, Intellektual'nyye i informatsionnyye sistemy. Intellekt-2021: Tr. Vseross. nauch.-tekhn. konf, Tula, Tula State University, 2021, pp. 183—189 (in Russian).
  21. Rashidov M. R. Development of an intelligent chat bot for advising applicants to DSU, Vseross. nauch.-prakt. konf. "Tsifrovaya transformatsiya v nauke, ekonomike i obrazovanii" v ramkakh realizatsii granta Minobrnauki Rossii "oSNOva nauki": sb. tezisov, Makhachkala, Dagestan State University, 2023, pp. 68—69 (in Russian).
  22. Gawade H., Patil V., Vishe P., Kolpe S. College Enquiry Chat-Bot System, International Journal of Engineering Research and Technology, 2020, vol. 9, no. 9, pp. 741—742.
  23. Punith S., Chaitra B., Veeranna K., Chethana R. M. Research Paper on Chatbot for Student Admission Enquiry, Journal of Advancement in Software Engineering and Testing, 2020, vol. 3, no. 1, pp. 1—9.
  24. Nesterova A. V., Kozhevyatkina M. A., Kechin E. S. Development of an intelligent chat bot for API NGT applicants, Tekhnologicheskiy suverenitet i tsifrovaya transformatsiya: Mezhdunar. nauch.-tekhn. konf. Kazan, Kazan State Energy University, 2024, pp. 158—161 (in Russian).
  25. Deeva O. V., Sharapov A. A. Development of an intelligent chat bot for SGUGIT applicants, Interekspo Geo-Sibir', 2020, vol. 7, no. 1, pp. 71—76. DOI: 10.33764/2618-981X-2020-7-1-71-76 (in Russian).
  26. GigaChat: documentation for developers, available at: https://developers.sber.ru/docs/ru/gigachat/models (date of access 25.02.2025).
  27. Gajbhiye A., Espinosa-Anke L., Schockaert S. Modelling Commonsense Properties using Pre-Trained Bi-Encoders, available at: https://arxiv.org/pdf/2210.02771 (date of access 25.02.2025).
  28. Malkov Yu. A., Yashunin D. A. Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs, available at: https://arxiv.org/abs/1603.09320 (date of access 25.02.2025).
  29. Reimers N., Gurevych I. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks, available at: https://arxiv.org/abs/1908.10084 (date of access 13.03.2025).