<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.3 20210610//EN" "JATS-journalpublishing1-3.dtd">
<article article-type="research-article" dtd-version="1.3" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xml:lang="ru"><front><journal-meta><journal-id journal-id-type="publisher-id">intechngu</journal-id><journal-title-group><journal-title xml:lang="ru">Вестник НГУ. Серия: Информационные технологии</journal-title><trans-title-group xml:lang="en"><trans-title>Vestnik NSU. Series: Information Technologies</trans-title></trans-title-group></journal-title-group><issn pub-type="ppub">1818-7900</issn><issn pub-type="epub">2410-0420</issn><publisher><publisher-name>НГУ</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="doi">10.25205/1818-7900-2024-22-4-5-16</article-id><article-id custom-type="elpub" pub-id-type="custom">intechngu-293</article-id><article-categories><subj-group subj-group-type="heading"><subject>Research Article</subject></subj-group><subj-group subj-group-type="section-heading" xml:lang="ru"><subject>Статьи</subject></subj-group></article-categories><title-group><article-title>О повышении качества выходных данных в вопросно-ответной системе, обрабатывающей климатическую информацию</article-title><trans-title-group xml:lang="en"><trans-title>On Increasing the Quality of the Climate Observations Question-Answering System’s Output Data</trans-title></trans-title-group></title-group><contrib-group><contrib contrib-type="author" corresp="yes"><contrib-id contrib-id-type="orcid">https://orcid.org/0000-0003-3619-1120</contrib-id><name-alternatives><name name-style="eastern" xml:lang="ru"><surname>Гавенко</surname><given-names>О. Ю.</given-names></name><name name-style="western" xml:lang="en"><surname>Gavenko</surname><given-names>O. Yu.</given-names></name></name-alternatives><bio xml:lang="ru"><p>Гавенко Ольга Юрьевна, доктор технических наук, кандидат филологических наук, ведущий научный сотрудник; старший преподаватель кафедры математического моделирования</p><p>Новосибирск</p></bio><bio xml:lang="en"><p>Olga Yu. Gavenko, Doctor of Sciences (Technical Sciences), Сandidate of Sciences (Philology), Leading Researcher; Senior lecturer of the Department of Mathematical Modeling</p><p>Novosibirsk</p></bio><email xlink:type="simple">olga.yu.gavenko@mail.ru</email><xref ref-type="aff" rid="aff-1"/></contrib><contrib contrib-type="author" corresp="yes"><contrib-id contrib-id-type="orcid">https://orcid.org/0009-0007-3658-6110</contrib-id><name-alternatives><name name-style="eastern" xml:lang="ru"><surname>Шашок</surname><given-names>Н. А.</given-names></name><name name-style="western" xml:lang="en"><surname>Shashok</surname><given-names>N. A.</given-names></name></name-alternatives><bio xml:lang="ru"><p>Шашок Наталья Александровна, аспирант</p><p>Новосибирск</p></bio><bio xml:lang="en"><p>Natalia A. Shashok, Ph. D Student</p><p>Novosibirsk</p></bio><email xlink:type="simple">n.shashok@alumni.nsu.ru</email><xref ref-type="aff" rid="aff-2"/></contrib></contrib-group><aff-alternatives id="aff-1"><aff xml:lang="ru"><institution>Федеральный исследовательский центр информационных и вычислительных технологий; Новосибирский государственный университет</institution><country>Россия</country></aff><aff xml:lang="en"><institution>Federal Research Center for Information and Computational Technologies; Novosibirsk State University</institution><country>Russian Federation</country></aff></aff-alternatives><aff-alternatives id="aff-2"><aff xml:lang="ru"><institution>Федеральный исследовательский центр информационных и вычислительных технологий</institution><country>Россия</country></aff><aff xml:lang="en"><institution>Federal Research Center for Information and Computational Technologies</institution><country>Russian Federation</country></aff></aff-alternatives><pub-date pub-type="collection"><year>2024</year></pub-date><pub-date pub-type="epub"><day>25</day><month>03</month><year>2025</year></pub-date><volume>22</volume><issue>4</issue><fpage>5</fpage><lpage>16</lpage><permissions><copyright-statement>Copyright &amp;#x00A9; Гавенко О.Ю., Шашок Н.А., 2025</copyright-statement><copyright-year>2025</copyright-year><copyright-holder xml:lang="ru">Гавенко О.Ю., Шашок Н.А.</copyright-holder><copyright-holder xml:lang="en">Gavenko O.Y., Shashok N.A.</copyright-holder><license xml:lang="ru" license-type="creative-commons-attribution" xlink:href="https://creativecommons.org/licenses/by/4.0/" xlink:type="simple"><license-p>Данная работа распространяется под лицензией Creative Commons Attribution 4.0.</license-p></license><license xml:lang="en" license-type="creative-commons-attribution" xlink:href="https://creativecommons.org/licenses/by/4.0/" xlink:type="simple"><license-p>This work is licensed under a Creative Commons Attribution 4.0 License.</license-p></license></permissions><self-uri xlink:href="https://intechngu.elpub.ru/jour/article/view/293">https://intechngu.elpub.ru/jour/article/view/293</self-uri><abstract><p>Разработка вопросно-ответной системы (QA), обрабатывающей климатическую информацию, опирается на использование разнородных климатических данных в различных форматах (текстовые, числовые, графические, видео, аудио, географические и данные мониторинга). Обязательным элементом вопросно-ответной системы должен являться инструмент, позволяющий обрабатывать и анализировать подобные данные.Процессы поиска и извлечения данных выступают центральной частью рассматриваемой системы, поскольку от них во многом зависит качество сгенерированного ответа. Точный способ извлечения данных имеет решающее значение для выходных данных системы QA, а также для проблем принятия решений, так как существуют ситуации, в которых LLM генерирует ответы, соответствующие контексту, но фактически являющиеся неверными и не соответствующими входным данным. Использование правильных метрик и алгоритмов для некоторых типов данных и неправильных для других может привести к превышению допустимого порога нерелевантных данных, что, в свою очередь, может снизить качество ответов. Дополненная поисковая генерация (Retrievalaugmented Generation, RAG) также может использоваться для оптимизации входных данных для этой задачи.В работе рассматриваются различные алгоритмы извлечения данных и ранжирования документов, а также возможность использования ансамблей агентов LLM при разработке вопросно-ответной системы, обрабатывающей климатическую информацию.</p></abstract><trans-abstract xml:lang="en"><p>The development of the climate observations question-answer (QA) information system relies on heterogeneous climate data in various formats (text, numerical, graphic, video, audio, geographic and monitoring data). A mandatory element of such a system is a tool that allows processing and analyzing such data.Searching and retrieving data is a central part of the system in question, since the quality of the generated answer heavily depends on it. The exact way the data is retrieved is critical to the output of a QA system as well as to decision-making problems, since there are situations in which the LLM generates a contextually appropriate but factually incorrect answers that do not match the input. Using correct metrics and algorithms for some data types and incorrect ones for others can cause the permissible threshold of irrelevant data to be exceeded, which in turn can cause the quality of the answers to decrease. Retrieval-augmented generation (RAG) systems can also be used to optimize input data for that task.This work discusses various algorithms for data extraction and document ranking, as well as the possibility of using ensembles of LLM agents in development of the QA system that works with climate data.</p></trans-abstract><kwd-group xml:lang="ru"><kwd>оптимизация входных данных</kwd><kwd>вопросно-ответные системы</kwd><kwd>разработка RAG-системы</kwd><kwd>обработка мультимодальных данных</kwd></kwd-group><kwd-group xml:lang="en"><kwd>Input data optimization</kwd><kwd>question answering systems</kwd><kwd>RAG system development</kwd><kwd>multimodal data processing</kwd></kwd-group></article-meta></front><back><ref-list><title>References</title><ref id="cit1"><label>1</label><citation-alternatives><mixed-citation xml:lang="ru">Hirschman L., Gaizauskas R. Natural language question answering: the view from here // Natural Language Engineering Journal. 2001. Vol. 7, no. 4. P. 275–300. DOI: 10.1017/S1351324901002807</mixed-citation><mixed-citation xml:lang="en">Hirschman L., Gaizauskas R. Natural language question answering: the view from here. Natural Language Engineering Journal, 2001, vol. 7, no. 4, pp. 275–300. DOI: 10.1017/S1351324901002807</mixed-citation></citation-alternatives></ref><ref id="cit2"><label>2</label><citation-alternatives><mixed-citation xml:lang="ru">Keen P. G. W, Michael S. S. M. Decision support systems: an organizational perspective. Michigan, Addison-Wesley, 1978.</mixed-citation><mixed-citation xml:lang="en">Keen P. G. W, Michael S. S. M. Decision support systems: an organizational perspective. Michigan, Addison-Wesley, 1978.</mixed-citation></citation-alternatives></ref><ref id="cit3"><label>3</label><citation-alternatives><mixed-citation xml:lang="ru">Woods W. A. Progress in natural language understanding: an application to lunar geology // Proceedings of the national computer conference and exposition (AFIPS ‘73), 1974, Association for Computing Machinery, New York, NY, USA, p. 441–450. DOI: https://doi.org/10.1145/1499586.1499695</mixed-citation><mixed-citation xml:lang="en">Woods W. A. Progress in natural language understanding: an application to lunar geology. Proceedings of the national computer conference and exposition (AFIPS ‘73), 1974, Association for Computing Machinery, New York, NY, USA, pp. 441–450. DOI: https://doi.org/10.1145/1499586.1499695</mixed-citation></citation-alternatives></ref><ref id="cit4"><label>4</label><citation-alternatives><mixed-citation xml:lang="ru">Lewis P., Perez E., et al. Retrieval-augmented generation for knowledge-intensive NLP tasks // Proceedings of the 34th International Conference on Neural Information Processing Systems (NIPS ‘20), 2020, Curran Associates Inc., Red Hook, NY, USA, Article 793, p. 9459–9474. DOI: 10.48550/arXiv.2005.11401</mixed-citation><mixed-citation xml:lang="en">Lewis P., Perez E., et al. Retrieval-augmented generation for knowledge-intensive NLP tasks. Proceedings of the 34th International Conference on Neural Information Processing Systems (NIPS ‘20), 2020, Curran Associates Inc., Red Hook, NY, USA, Article 793, pp. 9459–9474. DOI: 10.48550/arXiv.2005.11401</mixed-citation></citation-alternatives></ref><ref id="cit5"><label>5</label><citation-alternatives><mixed-citation xml:lang="ru">Wang L., Lo K. et al. CORD-19: The COVID-19 Open Research Dataset. ArXiv, abs/2004.10706, 2020. DOI: 10.48550/arXiv.2004.10706</mixed-citation><mixed-citation xml:lang="en">Wang L., Lo K. et al. CORD-19: The Covid-19 Open Research Dataset. ArXiv, abs/2004.10706, 2020. DOI: 10.48550/arXiv.2004.10706</mixed-citation></citation-alternatives></ref><ref id="cit6"><label>6</label><citation-alternatives><mixed-citation xml:lang="ru">Rajpurkar P., Zhang J., Lopyrev K., Liang P. Squad: 100,000+ questions for machine comprehension of text // Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, 2016, Association for Computational Linguistics, Austin, Texas, USA, p. 2383–2392. DOI: 10.18653/v1/D16-1264</mixed-citation><mixed-citation xml:lang="en">Rajpurkar P., Zhang J., Lopyrev K., Liang P. Squad: 100,000+ questions for machine comprehension of text. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, 2016, Association for Computational Linguistics, Austin, Texas, USA, pp. 2383–2392. doi: 10.18653/v1/D16-1264</mixed-citation></citation-alternatives></ref><ref id="cit7"><label>7</label><citation-alternatives><mixed-citation xml:lang="ru">Magesh V., Surani F., Dahl M., Suzgun M. et al. Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools. ArXiv, abs/2405.20362, 2024. DOI: 10.48550/arXiv.2405.20362</mixed-citation><mixed-citation xml:lang="en">Magesh V., Surani F., Dahl M., Suzgun M. et al. Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools. ArXiv, abs/2405.20362, 2024. DOI: 10.48550/arXiv.2405.20362</mixed-citation></citation-alternatives></ref><ref id="cit8"><label>8</label><citation-alternatives><mixed-citation xml:lang="ru">Page L., Brin S., Motwani R., Winograd T. The PageRank Citation Ranking: Bringing Order to the Web. Technical Report, Stanford InfoLab, 1999.</mixed-citation><mixed-citation xml:lang="en">Page L., Brin S., Motwani R., Winograd T. The PageRank Citation Ranking: Bringing Order to the Web. Technical Report, Stanford InfoLab, 1999.</mixed-citation></citation-alternatives></ref><ref id="cit9"><label>9</label><citation-alternatives><mixed-citation xml:lang="ru">Фадеев С. В. Экологический словарь. СПб., 2011.</mixed-citation><mixed-citation xml:lang="en">Fadeev S. V. Ekologicheskij slovar’. Saint Petersburg, 2011 (in Russ.)</mixed-citation></citation-alternatives></ref><ref id="cit10"><label>10</label><citation-alternatives><mixed-citation xml:lang="ru">Florin C., Giovanni T. et al: The Power of Noise: Redefining Retrieval for RAG Systems // Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, 2024, Association for Computing Machinery, New York, NY, USA, p. 719–729. DOI: 10.1145/3626772.3657834</mixed-citation><mixed-citation xml:lang="en">Florin C., Giovanni T., et al: The Power of Noise: Redefining Retrieval for RAG Systems. Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, 2024, Association for Computing Machinery, New York, NY, USA pp. 719-729. DOI: 10.1145/3626772.3657834</mixed-citation></citation-alternatives></ref><ref id="cit11"><label>11</label><citation-alternatives><mixed-citation xml:lang="ru">Cormack G. V., Clarke C. L., Büttcher S. Reciprocal rank fusion outperforms condorcet and individual rank learning methods // Proceedings of the 32nd international ACM SIGIR conference on Research and development in information retrieval, 2009, Association for Computing Machinery, New York, NY, USA, p. 758–759. DOI: 10.1145/1571941.1572114</mixed-citation><mixed-citation xml:lang="en">Cormack G. V., Clarke C. ., Büttcher S. Reciprocal rank fusion outperforms condorcet and individual rank learning methods. Proceedings of the 32nd international ACM SIGIR conference on Research and development in information retrieval, 2009, Association for Computing Machinery, New York, NY, USA, pp. 758–759. DOI: 10.1145/1571941.1572114</mixed-citation></citation-alternatives></ref></ref-list><fn-group><fn fn-type="conflict"><p>The authors declare that there are no conflicts of interest present.</p></fn></fn-group></back></article>
