ellibsЭлектронные библиотекиRussian Digital Libraries Journal1562-5419Казанский (Приволжский) федеральный университетellibs-719Research ArticleСтатьиЗапросы к нереляционным данным на естественном языке на основе большой языковой моделиQueries to Non-Relational Data using Natural Language based on a Large Language ModelЕркимбаевАдильбек ОмирбековичErkimbaevAdilbek Omirbekovichadilbek@jiht.ruЗицерманВладимир ЮрьевичZitsermanVladimir Yurievichvz1941@mail.ruКобзевГеоргий АнатольевичKobzevGeorge Anatolyevichgkbz@mail.ruОбъединенный институт высоких температур РАНJoint Institute for High Temperatures2026040320262917698Copyright © Еркимбаев А.О., Зицерман В.Ю., Кобзев Г.А., 20262026Еркимбаев А.О., Зицерман В.Ю., Кобзев Г.А.Erkimbaev A.O., Zitserman V.Y., Kobzev G.A.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://ellibs.elpub.ru/jour/article/view/719

В работе рассмотрены новые возможности организации запросов на естественном языке к научным локальным базам данных нереляционного типа. Проведенный анализ исследований, выполненных за последние годы, показал активное внедрение запросов на естественном языке к базам данных различного типа. Отмечено активное применение методов машинного обучения (нейронных алгоритмов). Показано широкое использование в последние два года большой языковой модели для подготовки запросов в различных языковых средах и областях знаний. Проведено исследование новых возможностей графовой базы данных AllegroGraph по использованию больших языковых моделей для организации поиска на естественном языке. Функционал базы данных изучен на примере системы метаданных по теплофизическим свойствам веществ в форме предметной онтологии «Термаль». Тестирование поисковых запросов в двуязычной (английская и русская) среде базы данных выявило в целом преодолимые проблемы и дает хорошие надежды на дальнейшее применение новых прикладных сервисов с использованием больших языковых моделей.

The main purpose of this work is to explore new opportunities for organizing natural language queries in scientific local databases that are not relational. A brief review of recent research shows that there has been an active introduction of natural language queries into databases of various types, and the use of machine learning methods, such as neural algorithms, is noted. The widespread use of large language models in the last two years for query generation in various language settings and fields of expertise has been demonstrated. A study has been conducted to explore the potential of the AllegroGraph graph database in using large language models for natural language search. The functionality of the database has been examined using the example of a metadata system for thermophysical properties in the form of the "Thermal" domain ontology. Testing search queries in a bilingual (English and Russian) database environment has revealed some general problems that can be overcome, and it gives us good hope for the future application of new services using large language models.

запрос на естественном языкебольшая языковая модельэмбеддингнереляционные базы данныхграфовая база данныхонтология предметной областиnatural language querylarge language modelembeddingnon-relational databasesgraph databasedomain ontologyReferencesErkimbaev A.O., Zitserman V.Iu., Kobzev G.A. Tipologiia materialovedcheskikh dannykh // Nauchno-tekhnicheskaia informatsiia. Ser. 2. 2023. № 6. S. 25–39.Erkimbaev A.O., Zitserman V.Iu., Kobzev G.A. Tipologiia materialovedcheskikh dannykh // Nauchno-tekhnicheskaia informatsiia. Ser. 2. 2023. № 6. S. 25–39.Erkimbaev A.O., Zitserman V.Iu., Kobzev G.A., Kosinov A.V. O predstavlenii i otsenke nauchnykh dannykh chislovogo i nechislovogo tipa pri provedenii issledovanii po svoistvam materialov // Nauchno-tekhnicheskaia informatsiia. Ser. 2. 2023. № 2. S. 8–16.Erkimbaev A.O., Zitserman V.Iu., Kobzev G.A., Kosinov A.V. O predstavlenii i otsenke nauchnykh dannykh chislovogo i nechislovogo tipa pri provedenii issledovanii po svoistvam materialov // Nauchno-tekhnicheskaia informatsiia. Ser. 2. 2023. № 2. S. 8–16.Woods W.A. Semantics and quantification in natural language question answering. // Advances in computers. N.Y. etc.: Acad. Press, 1978. Vol. 1 7. P. 1–87. URL: https://web.stanford.edu/class/linguist289/woods.pdfWoods W.A. Semantics and quantification in natural language question answering. // Advances in computers. N.Y. etc.: Acad. Press, 1978. Vol. 1 7. P. 1–87. URL: https://web.stanford.edu/class/linguist289/woods.pdfBorodin D.S., Stroganov Iu.V. K zadache sostavleniia zaprosov k bazam dannykh na estestvennom iazyke // Novye informatsionnye tekhnologii v avtomatizirovannykh sistemakh: materialy 19 nauchno-prakticheskogo seminara. M.: IPM im. M.V. Keldysha, aprel 2016. P. 119–125.Borodin D.S., Stroganov Iu.V. K zadache sostavleniia zaprosov k bazam dannykh na estestvennom iazyke // Novye informatsionnye tekhnologii v avtomatizirovannykh sistemakh: materialy 19 nauchno-prakticheskogo seminara. M.: IPM im. M.V. Keldysha, aprel 2016. P. 119–125.Bolshakova E.I., Klyshinskii E. S., Lande D.V., Noskov A.A., Peskova O.V., Iagunova E.V. Avtomaticheskaia obrabotka tekstov na estestvennom iazyke i kompiuternaia lingvistika: uchebnoe posobie. M.: MIEM, 2011. 272 s.Bolshakova E.I., Klyshinskii E. S., Lande D.V., Noskov A.A., Peskova O.V., Iagunova E.V. Avtomaticheskaia obrabotka tekstov na estestvennom iazyke i kompiuternaia lingvistika: uchebnoe posobie. M.: MIEM, 2011. 272 s.Borodin D.S., Stroganov Iu.V., Volkova L.L., Rudakov I.V., Prosukov E.A. Transliator zaprosov na ogranichennom estestvennom iazyke v zaprosy k reliatsionnym bazam dannykh // Sistemnyi administrator. 2019. Vypusk №01-02. S. 194–195.Borodin D.S., Stroganov Iu.V., Volkova L.L., Rudakov I.V., Prosukov E.A. Transliator zaprosov na ogranichennom estestvennom iazyke v zaprosy k reliatsionnym bazam dannykh // Sistemnyi administrator. 2019. Vypusk №01-02. S. 194–195.Posevkin R.V. Primenenie semanticheskoi modeli bazy dannykh pri realizatsii estestvenno-iazykovogo polzovatelskogo interfeisa // Nauchno-tekhnicheskii vestnik informatsionnykh tekhnologii, mekhaniki i optiki. 2018. Tom 18. № 2. S. 262–267.Posevkin R.V. Primenenie semanticheskoi modeli bazy dannykh pri realizatsii estestvenno-iazykovogo polzovatelskogo interfeisa // Nauchno-tekhnicheskii vestnik informatsionnykh tekhnologii, mekhaniki i optiki. 2018. Tom 18. № 2. S. 262–267.Mikolov T., et al. Distributed representations of words and phrases and their compositionality // Proc. 26th Int. Conf. on Neural Information Processing Systems. 2013. P. 3111–3119.Mikolov T., et al. Distributed representations of words and phrases and their compositionality // Proc. 26th Int. Conf. on Neural Information Processing Systems. 2013. P. 3111–3119.Pennington J., et al. Glove: Global vectors for word representation // Proc. Conf. Empirical Methods in Natural Language Processing. 2014. P. 1532–1543.Pennington J., et al. Glove: Global vectors for word representation // Proc. Conf. Empirical Methods in Natural Language Processing. 2014. P. 1532–1543.Kenton J.D.M.-W. C., Toutanova L.K. Bert: Pre-training of deep bidirectional transformers for language understanding // Proc. Conf. of North American Chapter of Association for Computational Linguistics. 2019. P. 4171–4186.Kenton J.D.M.-W. C., Toutanova L.K. Bert: Pre-training of deep bidirectional transformers for language understanding // Proc. Conf. of North American Chapter of Association for Computational Linguistics. 2019. P. 4171–4186.Hafsa Shareef Dar, M. Ikramullah Lali, Khalid Mahmood Malik, Syed Ahmad Chan Bukhari. Frameworks for Querying Databases Using Natural Language: A Literature Review. 2019. P. 1–18. arXiv preprint. URL: https://arxiv.org/abs/1909.01822Hafsa Shareef Dar, M. Ikramullah Lali, Khalid Mahmood Malik, Syed Ahmad Chan Bukhari. Frameworks for Querying Databases Using Natural Language: A Literature Review. 2019. P. 1–18. arXiv preprint. URL: https://arxiv.org/abs/1909.01822Baig Muhammad Shahzaib, et al. Natural Language to SQL Queries: A Review Original Article // International Journal of Innovations in Science & Technology. 2022. Vol. 4. Issue 1. P. 147–162.Baig Muhammad Shahzaib, et al. Natural Language to SQL Queries: A Review Original Article // International Journal of Innovations in Science & Technology. 2022. Vol. 4. Issue 1. P. 147–162.Tao Yu, et al. Spider: A large-scale human-labeled dataset for complex and cross-domain semantic parsing and text-to-SQL task. arXiv preprint. 2018. URL: https://arxiv.org/abs/1809.08887Tao Yu, et al. Spider: A large-scale human-labeled dataset for complex and cross-domain semantic parsing and text-to-SQL task. arXiv preprint. 2018. URL: https://arxiv.org/abs/1809.08887Manning C.D. Human language understanding & reasoning // Daedalus 2022. Vol. 151. Issue 2. P. 127–138.Manning C.D. Human language understanding & reasoning // Daedalus 2022. Vol. 151. Issue 2. P. 127–138.Meyer Jesse G., et al. ChatGPT and large language models in academia: opportunities and challenges // BioData Mining 2023. Vol. 16. Art. numb. 20.Meyer Jesse G., et al. ChatGPT and large language models in academia: opportunities and challenges // BioData Mining 2023. Vol. 16. Art. numb. 20.Microsoft Copilot в Azure с базой данных SQL Azure. URL: https://learn.microsoft.com/ru-ru/azure/azure-sql/copilot/copilot-azure-sql-overview?view=azuresqlMicrosoft Copilot в Azure с базой данных SQL Azure. URL: https://learn.microsoft.com/ru-ru/azure/azure-sql/copilot/copilot-azure-sql-overview?view=azuresqlMongoDB Query Generator using OpenAI. URL: https://www.mongodb.com/docs/compass/current/query-with-natural-language/#std-label-compass-query-natural-languageMongoDB Query Generator using OpenAI. URL: https://www.mongodb.com/docs/compass/current/query-with-natural-language/#std-label-compass-query-natural-languageLower your Large Language Model costs with Graphwise GraphDB. URL: https://www.ontotext.com/blog/lower-your-llm-costs-with-graphwise-graphdb/Lower your Large Language Model costs with Graphwise GraphDB. URL: https://www.ontotext.com/blog/lower-your-llm-costs-with-graphwise-graphdb/AllegroGraph 8.4.0 LLM Embed Specification. URL: https://franz.com/agraph/support/documentation/llmembed.htmlAllegroGraph 8.4.0 LLM Embed Specification. URL: https://franz.com/agraph/support/documentation/llmembed.htmlStardog Voicebox FAQ: How LLM, Generative AI, and Knowledge Graphs are the Future of Data Management. URL: https://www.stardog.com/blog/stardog-voicebox-faq-how-llm-generative-ai-and-knowledge-graphs-are-the-future-of-data-management/Stardog Voicebox FAQ: How LLM, Generative AI, and Knowledge Graphs are the Future of Data Management. URL: https://www.stardog.com/blog/stardog-voicebox-faq-how-llm-generative-ai-and-knowledge-graphs-are-the-future-of-data-management/Trakhtengerts M.S. Tekhnologiia podgotovki informatsii dlia baz dannykh v obmennom formate ISO 2709 // Nauchno-tekhnicheskaia informatsiia. Ser. 2. 2006. № 7. S. 28–31.Trakhtengerts M.S. Tekhnologiia podgotovki informatsii dlia baz dannykh v obmennom formate ISO 2709 // Nauchno-tekhnicheskaia informatsiia. Ser. 2. 2006. № 7. S. 28–31.

The authors declare that there are no conflicts of interest present.