Análisis de patrones predictivos en litigios administrativos y tributarios mediante IA generativa y expresiones regulares
DOI:
https://doi.org/10.47975/ijdl.v7.1305Palabras clave:
Modelos de lenguaje grandes; clasificación de documentos; expresiones regulares; análisis predictivo de patrones; IA generativa.Resumen
La clasificación de texto consiste en asignar un fragmento de texto a una categoría adecuada. El conjunto de categorías posibles varía según el dominio. En este estudio, evaluamos la precisión de ChatGPT al clasificar sentencias judiciales emitidas por un tribunal argentino mediante indicaciones basadas en coincidencias de expresiones regulares. Comparamos su rendimiento con dos enfoques alternativos: (1) clasificación basada en descripciones de las categorías en lenguaje natural y (2) un algoritmo tradicional basado en expresiones regulares. Los patrones de expresiones regulares fueron desarrollados por expertos del sector de la organización asociada, y todos los experimentos se realizaron en español. Nuestros resultados muestran que ChatGPT alcanza una alta precisión cuando se le solicitan descripciones en lenguaje natural e identificaciones de categorías basadas en expresiones regulares. Sin embargo, generar resultados con ChatGPT requiere mucho más tiempo. A pesar de esta limitación, la interfaz accesible del modelo y sus mínimos requisitos técnicos lo convierten en una herramienta prometedora para automatizar la clasificación de documentos. Además, el uso de indicaciones basadas en expresiones regulares permite la supervisión humana y la participación activa en el proceso de clasificación, una ventaja en dominios donde la interpretabilidad y la rendición de cuentas son cruciales.
Descargas
Citas
ABUKAR AHMED, Abdinor; KHALIF ALI, Mohamed. User-centric adoption of democratized generative AI: Focus on human-machine interaction and overcoming challenges. International Journal of Engineering Trends and Technology, v. 72, issue 9, p. 78-95, 2024.
AGARWAL, Basant; MITTAL, Namita. Text classification using machine learning methods-a survey. In: Proceedings of the Second International Conference on Soft Computing for Problem Solving (SocProS 2012). New Dehli: Springer, 2014. p. 701-709.
BHAMBRI, Gaurav. Information overload in business organizations and entrepreneurship: An analytical review of the literature. Business Information Review, v. 38, issue 4, 2021, p. 193–200.
BIDERMAN, Stella Biderman, et al. Lessons from the trenches on reproducible evaluation of language models. 2024. Available at: <https://arxiv.org/abs/2405.14782>.; REUEL, Anka; et al. Open problems in technical AI governance. 2024. Available at: <https://arxiv.org/abs/2407.14981>.
BURDEN, John. Evaluating AI evaluation: perils and prospects. 2024. Available at: <https://arxiv.org/abs/2407.09221>.
CABELLERO, William N.; JENKINS, Phillip R. On large language models in national security applications. Stat, v. 14, issue 2, e70057, 2025.
FRONTIER MODEL FORUM. Issue brief: Early best practices for frontier ai safety evaluations. 2024. Available at: <https://www.frontiermodelforum.org/updates/early-best-practices-for-frontier-ai-safety-evaluations/>.
JIANG, Shuo; HU, Jie; MAGEE, Christopher L.; LUO, Jianxi. Deep learning for technical document classification. IEEE Transactions on Engineering Management, v. 71, p. 1163-1179, 2022.
KLEIN, L. K; EARL, E.; CUNDICK, D. Reducing information overload in your organization. Harvard Business Review, 2023.
KOSTINA, Arina; DIKAIAKOS, Marios D.; STEFANIDIS, Dimosthenis; PALLIS, George Pallis. Large language models for text classification: case study and comprehensive review. 2025. Available at: <https://arxiv.org/abs/2501.08457>.
KOWSARI, Kamram; JAFARI MEIMANDI, Kiana; HEIDARYSAFA, Mojtaba; MENDU, Sanjana; BARNES, Laura; BROWN, Donald. Text classification algorithms: A survey. Information, v. 10, issue 4, p. 150, 2019.
LI, Yinghao; RAMPRASAD, Rampi; ZHANG, Chao. A simple but effective approach to improve structured language model output for information extraction. 2024. Available at: <https://arxiv.org/abs/2402.13364>.
MAHONEY, Christian; GRONVALL, Peter; HUBER-FLIFLET; ZHANG, Jianping. Explainable text classification techniques in legal document review: locating rationales without using human annotated training text snippets. In: 2022 IEEE International Conference on Big Data (Big Data). IEEE, p. 2044-2051, 2022.
MCINTOSH, Timothy R., et al. Inadequacies of large language model benchmarks in the era of generative artificial intelligence. IEEE Transactions on Artificial Intelligence, 2025. Available at: <https://arxiv.org/abs/2402.09880>.
NATIONAL JUDICIAL STATISTICS SYSTEM (SNEJ). Annual Report: Technical Report. National Judicial Statistics System (SNEJ): Argentina, 2022.
PALANIVINAYAGAM, Ashokkumar; ZIAD EL-BAYEH, Claude; DAMAŠEVICIUS, Robertas. Twenty years of machine-learning-based text classification: A systematic review. Algorithms, v. 16, issue 5, p. 236, 2023.
PASKOV, Patricia; BERGLUND, Lukas; SMITH, Everett; SODER, Lisa. GPAI evaluations standards taskforce: towards effective AI governance. 2024. Available at: <https://arxiv.org/abs/2411.13808>.
PEÑA ALMANSA, Alejandro; et al. Leveraging large language models for topic classification in the domain of public affairs. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2023.
PEREZ, Ethan; et al. Discovering language model behaviors with model-written evaluations. In: Findings of the Association for Computational Linguistics. Toronto: ACL, 2023. p. 13387–13434.
ROHERA, Pritika; GINIMAV, Chaitrali; SAWANT, Gayatri; JOSHI, Raviraj. Better to ask in english? Evaluating factual accuracy of multilingual LLMms in english and low-resource languages. Available at: <https://arxiv.org/abs/2504.20022>.
SAHOO, Pranab; SINGH, Ayush Kumar; SAHA, Sriparna; JAIN, Vinija; MONDAL, Samrat; CHADHA, Aman. A systematic survey of prompt engineering in large language models: techniques and applications. 2024. Available at: <https://arxiv.org/abs 2402.07927>.
SCHUT, Lisa; GAL, Yarin; FARQUHAR, Sebastian. Do multilingual LLMs think in english? 2025. Available at: <https://arxiv.org/abs/2502.15603>.
SCLAR, Melanie; CHOI, Yejin; TSVETKOV, Yulia; SUHR, Alane. Quantifying language models’ sensitivity to spurious features in prompt design or: how I learned to start worrying about prompt formatting. 2023. Available at: <https://arxiv.org/abs/2310.11324>.
SEGER, Elizabeth. What do we mean when we talk about “AI democratisation”. 2023. Available at: <https://www.governance.ai/analysis/what-do-we-mean-when-we-talk-about-ai-democratisation>.
SUN, Xiaofei; LI, Xiaoya; LI, Jiwei; WU, Fei; GUO, Shangwei; ZHANG, Tianwei; WANG, Guoyin. Text classification via large language models. 2023. Available at: <https://arxiv.org/abs/2305.08377>.
SUPREME COURT OF ARGENTINA. Statistics For The Year 2024: Technical report. Supreme Court of Argentina, 2024.
WAGH, Vedangi; KHANDVE, Snehal; JOSHI, Isha; WANI, Apurva; KALE, Geetanjali; JOSHI, Raviraj. Comparative study of long document classification. In: TENCON 2021-2021 IEEE Region 10 Conference (TENCON). 2021. p. 732–737.
WEIDINGER, Laura; et al. Sociotechnical safety evaluation of generative AI systems. 2023. Available at: <https://arxiv.org/abs/2310.11986>.
YE, Junjie; XU, Nuo; WANG, Yikun; ZHOU, Jie; ZHANG, Qi; GUI, Tao; HUANG, Xuanjing. LLM-DA: Data augmentation via large language models for few-shot named entity recognition. 2024. Available at: <https://arxiv.org/abs/2402.14568>.
YOON, Sungwook Yoon; et. al. Digital innovation in public administration through intelligent public sector automation (IPSA): Strategies and challenges. Journal of Multimedia Information System, v. 11, issue 4, p. 249-260, 2024.
ZHANG, Shuai; GU, Xiaodong; CHEN, Yuting; SHEN, Beijun. Infere: step-by-step regex generation via chain of inference. In: 38th IEEE/ACM International Conference on Automated Software Engineering (ASE). IEEE, 2023. p. 1505–1515.
Descargas
Publicado
Cómo citar
Licencia
Derechos de autor 2026 Juan Gustavo Corvalán, Sofía Tammaro, Gisel Alvarado, Luca Nicolás Forziati Gangi, Carina Mariel Papini, Mariana Sánchez Caparrós, Agostina Celeste Jara Rey, Florencia Paola Croci, Lola Ramos Pereyra, Melania Gadea, Luciano Dalla Via (Autor)
Esta obra está bajo una licencia internacional Creative Commons Atribución 4.0.
Autores que publicam nesta revista concordam com os seguintes termos:
- Autores mantém os direitos autorais e concedem à revista o direito de primeira publicação, com o trabalho simultaneamente licenciado sob a Creative Commons - Atribuição 4.0 Internacional que permite o compartilhamento do trabalho com reconhecimento da autoria e publicação inicial nesta revista.
- Autores têm autorização para assumir contratos adicionais separadamente, para distribuição não-exclusiva da versão do trabalho publicada nesta revista (ex.: publicar em repositório institucional ou como capítulo de livro), com reconhecimento de autoria e publicação inicial nesta revista.
- Autores têm permissão e são estimulados a publicar e distribuir seu trabalho online (ex.: em repositórios institucionais ou na sua página pessoal) a qualquer ponto antes ou durante o processo editorial, já que isso pode gerar alterações produtivas, bem como aumentar o impacto e a citação do trabalho publicado (Veja O Efeito do Acesso Livre).
