Ítem
Acceso Abierto

Integración del aprendizaje por refuerzo en parsers semánticos para la deducción lógica en lenguaje natural

Mostrar el registro sencillo de la publicación
dc.contributor.advisorAndrade Lotero, Edgar José
dc.creatorGómez Vargas, Camilo Andrés
dc.creator.degreeMagíster en Matemáticas Aplicadas y Ciencias de la Computación
dc.date.accessioned2025-09-22T20:35:28Z
dc.date.available2025-09-22T20:35:28Z
dc.date.created2025-08-27
dc.descriptionEl procesamiento del lenguaje natural (NLP) es una subdisciplina de la inteligencia artificial centrada en la interacción entre las computadoras y los seres humanos mediante lenguaje natural. Su objetivo es desarrollar modelos y sistemas que comprendan, interpreten y generen lenguaje natural de manera similar a como lo hace una persona. En este contexto, los parsers semánticos juegan un papel fundamental, ya que son herramientas que descomponen y representan la estructura y el significado de las oraciones. Estos permiten transformar el texto en una representación formal, proporcionando un medio para que los sistemas inteligentes cuenten con una representación del significado subyacente de las palabras y relaciones. Aunque los parsers semánticos son herramientas de gran importancia en NLP, estos métodos tienden a depender de reglas preestablecidas o de modelos supervisados que aprenden de ejemplos etiquetados, limitando su capacidad para la generalización y representación de nuevas estructuras. Esta falta de flexibilidad de los parsers para adaptarse a nuevas oraciones o a estructuras más complejas sin la necesidad de re-entrenamiento o de una definición más amplia de reglas gramaticales, restringen su utilidad en tareas complejas de inferencia y razonamiento lógico. Por tanto, el objeto de estudio de este trabajo es desarrollar un sistema que utilice aprendizaje por refuerzo profundo para optimizar la representación de estructuras lógicas a partir de oraciones en lenguaje natural. Así, esta investigación desarrolla un modelo capaz de realizar representaciones de silogismos con estructuras conjuntivas e implicatorias. El trabajo se centra en la definición del entorno de aprendizaje, la señal de recompensas, el esquema de entrenamiento y la evaluación de resultados. De esta manera, se busca mejorar la capacidad de las máquinas para interpretar y razonar sobre el lenguaje, lo cual representa un avance en el desarrollo de sistemas de inteligencia artificial que puedan operar con un razonamiento estructurado, consistente y fundamentado.
dc.description.abstractNatural language processing (NLP) is an artificial intelligence subdiscipline focused on the interaction between computers and humans through natural language. The goal of NLP is to develop models and systems that can understand, interpret, and generate natural language like a human. Semantic parsers play a fundamental role in this context because they break down and represent the structure and meaning of sentences. Semantic parsers transform text into a formal representation, providing intelligent systems with a means to understand the underlying meaning of words and relationships. However, these methods tend to rely on pre-established rules or supervised models that learn from labeled examples, which limits their ability to generalize and represent new structures. The inability of parsers to adapt to new sentences or more complex structures without retraining or a broader definition of grammatical rules restricts their usefulness in complex inference and logical reasoning tasks. Thus, this study aims to develop a system that uses deep reinforcement learning to optimize the representation of logical structures from sentences in natural language. This research develops a model capable of representing syllogisms with conjunctive and implicative structures. The study focuses on defining the learning environment, reward signal, training scheme, and evaluation of results. This approach aims to enhance machines' ability to interpret and reason about language, representing an advancement in the development of artificial intelligence systems capable of operating with structured, consistent, and well-founded reasoning.
dc.format.extent58 pp
dc.format.mimetypeapplication/pdf
dc.identifier.doihttps://doi.org/10.48713/10336_46569
dc.identifier.urihttps://repository.urosario.edu.co/handle/10336/46569
dc.language.isospa
dc.publisherUniversidad del Rosariospa
dc.publisher.departmentEscuela de Ciencias e Ingenieríaspa
dc.publisher.programMaestría en Matemáticas Aplicadas y Ciencias de la Computaciónspa
dc.rightsAttribution 4.0 International*
dc.rights.accesRightsinfo:eu-repo/semantics/openAccess
dc.rights.accesoAbierto (Texto Completo)
dc.rights.urihttp://creativecommons.org/licenses/by/4.0/*
dc.source.bibliographicCitationLu, Xuantao; Liu, Jingping; Gu, Zhouhong; Tong, Hanwen; Xie, Chenhao; Huang, Junyang; Xiao, Yanghua; Wang, Wenguang; Calzolari, Nicoletta; Huang, Chu-Ren; Kim, Hansaem; Pustejovsky, James; Wanner, Leo; Choi, Key-Sun; Ryu, Pum-Mo; Chen, Hsin-Hsi; Donatelli, Lucia; Ji, Heng; Kurohashi, Sadao; Paggio, Patrizia; Xue, Nianwen; Kim, Seokhwan; Hahm, Younggyun; He, Zhong; Lee, Tony Kyungil; Santus, Enrico; Bond, Francis; Na, Seung-Hoon (2022) Parsing Natural Language into Propositional and First-Order Logic with. En: Proceedings of the 29th International Conference on Computational. pp. 5419-5431 : International Committee on Computational Linguistics; Disponible en: https://aclanthology.org/2022.coling-1.481/.
dc.source.bibliographicCitationHan, Simeng; Schoelkopf, Hailey; Zhao, Yilun; Qi, Zhenting; Zhou, Martin Riddell Wenfei; Coady, James; Peng, David; Qiao, Yujie; Benson, Luke; Sun, Lucy; Szabo, Alex Wardle-Solano Hannah; Zubova, Ekaterina; Burtell, Matthew; Fan, Jonathan; Liu, Yixin; Wong, Brian; Sailor, Malcolm; Ni, Ansong; Nan, Linyong; Kasai, Jungo; Yu, Tao; Zhang, Rui; Fabbri, Alexander R; Kryscinski, Wojciech; Yavuz, Semih; Liu, Ye; Lin, Xi Victoria; Joty, Shafiq; Zhou, Yingbo; Xiong, Caiming; Ying, Rex; Cohan, Arman; Radev, Dragomir (2022) FOLIO: Natural Language Reasoning with First-Order Logic. Disponible en: https://arxiv.org/abs/2209.00840.
dc.source.bibliographicCitationKamp, Hans; Reyle, Uwe (1993) From Discourse to Logic: Introduction to Modeltheoretic Semantics of. Dordrecht, The Netherlands: Springer;
dc.source.bibliographicCitationEisenstein, Jacob (2019) Introduction to Natural Language Processing. : MIT Press;
dc.source.bibliographicCitationSutton, Richard S; Barto, Andrew G (2018) Reinforcement Learning, second edition: An Introduction. Cambridge, MA: MIT Press;
dc.source.bibliographicCitationJang, Beakcheol; Kim, Myeonghwi; Harerimana, Gaspard; Kim, Jong Wook (2019) Q-Learning Algorithms: A Comprehensive Classification and Applications. En: IEEE Access. Vol. 7; pp. 133653-133667 Disponible en: http://dx.doi.org/10.1109/ACCESS.2019.2941229. Disponible en: 10.1109/ACCESS.2019.2941229.
dc.source.bibliographicCitationMnih, Volodymyr; Kavukcuoglu, Koray; Silver, David; Graves, Alex; Antonoglou, Ioannis; Wierstra, Daan; Riedmiller, Martin (2013) Playing Atari with Deep Reinforcement Learning. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/1312.5602.
dc.source.bibliographicCitationIBM (2024) Large language models. 2024/11/1; Consultado en: 2024/11/1. Disponible en: https://www.ibm.com/topics/large-language-models.
dc.source.bibliographicCitationUrrutia, Daniel (2024) Qué es un analizador sintáctico o parser | Definición, significado y para. Disponible en: https://www.arimetrics.com/glosario-digital/analizador-sintactico-parser.
dc.source.bibliographicCitationLi, Yuchen; Risteski, Andrej (2021) The Limitations of Limited Context for Constituency Parsing. Disponible en: https://arxiv.org/abs/2106.01580.
dc.source.bibliographicCitationUc-Cetina, Víctor; Navarro-Guerrero, Nicolás; Martin-Gonzalez, Alejandro; Weber, Christian; Wermter, Stefan (2022) Survey on reinforcement learning for language processing. En: Artificial Intelligence Review. Vol. 56; No. 2; pp. 1543-1575 Disponible en: http://dx.doi.org/10.1007/s10462-022-10205-5. Disponible en: 10.1007/s10462-022-10205-5.
dc.source.bibliographicCitationWu, Zhaofeng; Qiu, Linlu; Ross, Alexis; Akyürek, Ekin; Chen, Boyuan; Wang, Bailin; Kim, Najoung; Andreas, Jacob; Kim, Yoon (2023) Reasoning or Reciting? Exploring the Capabilities and Limitations of. Disponible en: https://arxiv.org/abs/2307.02477.
dc.source.bibliographicCitationFrançois-Lavet, Vincent; Henderson, Pierre; Islam, Rashed; Bellemare, Marc G; Pineau, Joelle (2018) An Introduction to Deep Reinforcement Learning. En: Foundations and Trends® in Machine Learning. Vol. 11; No. 3-4; pp. 219-354 Disponible en: http://dx.doi.org/10.1561/2200000071. Disponible en: 10.1561/2200000071.
dc.source.bibliographicCitationDoan, Anhai; Halevy, Alon; Ives, Zachary; Doan, Anhai; Halevy, Alon; Ives, Zachary (2012) 4. En: Principles of Data Integration. pp. 95-119 Boston: Morgan Kaufmann; 9780124160446; Disponible en: http://dx.doi.org/10.1016/B978-0-12-416044-6.00004-1; https://www.sciencedirect.com/science/article/pii/B9780124160446000041. Disponible en: 10.1016/B978-0-12-416044-6.00004-1.
dc.source.bibliographicCitationKamath, Aishwarya; Das, Rajarshi (2019) A Survey on Semantic Parsing. En: Automated Knowledge Base Construction (AKBC). Disponible en: https://openreview.net/forum?id=HylaEWcTT7.
dc.source.bibliographicCitationPiza-Londono, Jessenia; Andrade-Lotero, Edgar J; Orjuela-Canon, Alvaro David (2024) Leveraging Semantic Parsing using Text Embeddings and Reinforcement. En: 2024 IEEE Latin American Conference on Computational Intelligence, LA-CCI.: Institute of Electrical and Electronics Engineers Inc.; Disponible en: http://dx.doi.org/10.1109/LA-CCI62337.2024.10814850. Disponible en: 10.1109/LA-CCI62337.2024.10814850.
dc.source.bibliographicCitationVaswani, Ashish; Shazeer, Noam; Parmar, Niki; Uszkoreit, Jakob; Jones, Llion; Gomez, Aidan N; Kaiser, Lukasz; Polosukhin, Illia (2023) Attention Is All You Need. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/1706.03762.
dc.source.bibliographicCitationZheng, Zifan; Wang, Yezhaohui; Huang, Yuxin; Song, Shichao; Yang, Mingchuan; Tang, Bo; Xiong, Feiyu; Li, Zhiyu (2024) Attention Heads of Large Language Models: A Survey. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/2409.03752.
dc.source.bibliographicCitationKeneshloo, Yaser; Shi, Tian; Ramakrishnan, Naren; Reddy, Chandan K (2019) Deep Reinforcement Learning For Sequence to Sequence Models. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/1805.09461.
dc.source.bibliographicCitationBengio, Yoshua; Louradour, Jérôme; Collobert, Ronan; Weston, Jason (2009) Curriculum learning. En: Proceedings of the 26th Annual International Conference on Machine. pp. 41–48 : Association for Computing Machinery; Disponible en: https://doi.org/10.1145/1553374.1553380; http://dx.doi.org/10.1145/1553374.1553380. Disponible en: 10.1145/1553374.1553380.
dc.source.bibliographicCitationSoviany, Petru; Ionescu, Radu Tudor; Rota, Paolo; Sebe, Nicu (2022) Curriculum Learning: A Survey. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/2101.10382.
dc.source.bibliographicCitationMatiisen, Tambet; Oliver, Avital; Cohen, Taco; Schulman, John (2017) Teacher-Student Curriculum Learning. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/1707.00183.
dc.source.bibliographicCitationJiang, Lu; Meng, Deyu; Zhao, Qian; Shan, Shiguang; Hauptmann, Alexander (2015) Self-paced Curriculum Learning. En: No.1: The Twenty-Ninth Conference on Artificial Intelligence Volume. pp. 2694-2695
dc.source.bibliographicCitation Gabbay, D M; Guenthner, F (2007) Handbook of Philosophical Logic. : Springer Finance;
dc.source.bibliographicCitationSchaul, Tom; Quan, John; Antonoglou, Ioannis; Silver, David (2016) Prioritized Experience Replay. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/1511.05952.
dc.source.bibliographicCitationBekki, Daisuke; Onada, Takashi; Bekki, Daisuke; McCready, Elin (2011) Combinatory Categorial Grammar as a Substructural Logic. En: New Frontiers in Artificial Intelligence. pp. 16-29 : Springer Berlin Heidelberg;
dc.source.bibliographicCitationLiang, Chen; Berant, Jonathan; Le, Quoc; Forbus, Kenneth D; Lao, Ni (2017) Neural Symbolic Machines: Learning Semantic Parsers on Freebase with Weak. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/1611.00020.
dc.source.bibliographicCitationZhou, Yujun; Ye, Jiayi; Ling, Zipeng; Han, Yufei; Huang, Yue; Zhuang, Haomin; Liang, Zhenwen; Guo, Kehan; Guo, Taicheng; Wang, Xiangqi; Zhang, Xiangliang (2025) Dissecting Logical Reasoning in LLMs: A Fine-Grained Evaluation and. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/2506.04810.
dc.source.bibliographicCitationPan, Liangming; Albalak, Alon; Wang, Xinyi; Wang, William Yang (2023) Logic-LM: Empowering Large Language Models with Symbolic Solvers for. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/2305.12295.
dc.source.bibliographicCitationCs224n, Stanford; Project, Custom; Elezabi, Hashem; Akoush, Bassem (2025) Prototype-then-Refine: A Neurosymbolic Approach for Improved Logical. En: Stanford CS224N. Disponible en: https://api.semanticscholar.org/CorpusID:269325536.
dc.source.bibliographicCitationShojaee*†, Parshin; Mirzadeh*, Iman; Alizadeh, Keivan; Horton, Maxwell; Bengio, Samy; Farajtabar, Mehrdad (2025) The Illusion of Thinking: Understanding the Strengths and Limitations of. Disponible en: https://ml-site.cdn-apple.com/papers/the-illusion-of-thinking.pdf.
dc.source.bibliographicCitationYao, Shunyu; Yu, Dian; Zhao, Jeffrey; Shafran, Izhak; Griffiths, Thomas L; Cao, Yuan; Narasimhan, Karthik (2023) Tree of Thoughts: Deliberate Problem Solving with Large Language Models. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/2305.10601.
dc.source.bibliographicCitationHe, Tao; Li, Hao; Chen, Jingchang; Liu, Runxuan; Cao, Yixin; Liao, Lizi; Zheng, Zihao; Chu, Zheng; Liang, Jiafeng; Liu, Ming (2025) A Survey on Complex Reasoning of Large Language Models through the Lens of. Disponible en: http://dx.doi.org/10.13140/RG.2.2.23943.30886. Disponible en: 10.13140/RG.2.2.23943.30886.
dc.source.bibliographicCitationChen, Qiguang; Qin, Libo; Liu, Jinhao; Peng, Dengyun; Guan, Jiannan; Wang, Peng; Hu, Mengkang; Zhou, Yuhang; Gao, Te; Che, Wanxiang (2025) Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning. En: arXiv [cs.AI]. Disponible en: http://arxiv.org/abs/2503.09567.
dc.source.bibliographicCitationLee, Jinu; Hockenmaier, Julia (2025) Evaluating Step-by-step Reasoning Traces: A Survey. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/2502.12289.
dc.source.bibliographicCitationMirzadeh, Iman; Alizadeh, Keivan; Shahrokhi, Hooman; Tuzel, Oncel; Bengio, Samy; Farajtabar, Mehrdad (2024) GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/2410.05229.
dc.source.bibliographicCitationHuang, Lei; Yu, Weijiang; Ma, Weitao; Zhong, Weihong; Feng, Zhangyin; Wang, Haotian; Chen, Qianglong; Peng, Weihua; Feng, Xiaocheng; Qin, Bing; Liu, Ting (2025) A Survey on Hallucination in Large Language Models: Principles, Taxonomy,. En: ACM Transactions on Information Systems. Vol. 43; No. 2; pp. 1–55 : Association for Computing Machinery (ACM); 1558-2868; Disponible en: http://dx.doi.org/10.1145/3703155. Disponible en: 10.1145/3703155.
dc.source.bibliographicCitationGamut, L T F; Andrade-Lotero, Edgar (translators), Carlos Márquez (2010) Lógica, lenguaje y significado: lógica intensional y gramática lógica. Bogotá: Editorial Universidad del Rosario;
dc.source.bibliographicCitationJurafsky, Daniel; Martin, James (2008) Speech and Language Processing. : Prentice Hall;
dc.source.bibliographicCitationChang, Yupeng; Wang, Xu; Wang, Jindong; Wu, Yuan; Yang, Linyi; Zhu, Kaijie; Chen, Hao; Yi, Xiaoyuan; Wang, Cunxiang; Wang, Yidong; Ye, Wei; Zhang, Yue; Chang, Yi; Yu, Philip S; Yang, Qiang; Xie, Xing (2024) A Survey on Evaluation of Large Language Models. En: ACM Trans. Intell. Syst. Technol. Vol. 15; No. 3; New York, NY, USA: Association for Computing Machinery; 2157-6904; Disponible en: https://doi.org/10.1145/3641289; http://dx.doi.org/10.1145/3641289. Disponible en: 10.1145/3641289.
dc.source.bibliographicCitationXu, Fangzhi; Lin, Qika; Han, Jiawei; Zhao, Tianzhe; Liu, Jun; Cambria, Erik (2024) Are Large Language Models Really Good Logical Reasoners? A Comprehensive. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/2306.09841.
dc.source.bibliographicCitationMcDermott, D V (1987) A critique of pure reason. En: Computational Intelligence. Vol. 3; pp. 151-160
dc.source.bibliographicCitationPaulus, Romain; Xiong, Caiming; Socher, Richard (2017) A Deep Reinforced Model for Abstractive Summarization. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/1705.04304.
dc.source.bibliographicCitationWiewiora, E (2003) Potential-Based Shaping and Q-Value Initialization are Equivalent. En: Journal of Artificial Intelligence Research. Vol. 19; pp. 205–208 : AI Access Foundation; 1076-9757; Disponible en: http://dx.doi.org/10.1613/jair.1190. Disponible en: 10.1613/jair.1190.
dc.source.bibliographicCitationOzdemir, Sinan (2024) Quick Start Guide to Large Language Models: Strategies and Best Practices. : Addison-Wesley Professional; 9780135346570;
dc.source.bibliographicCitationMinsky, Marvin; Winston, P H (1975) A framework for representing knowledge. En: The Psychology of Computer Vision.: McGraw-Hill;
dc.source.bibliographicCitationBasmov, Victoria; Goldberg, Yoav; Tsarfaty, Reut (2023) Simple Linguistic Inferences of Large Language Models (LLMs): Blind Spots. Disponible en: https://arxiv.org/abs/2305.14785.
dc.source.bibliographicCitationGrissom, Alvin, II; He, He; Boyd-Graber, Jordan; Morgan, John; Daumé, Hal, III; Moschitti, Alessandro; Pang, Bo; Daelemans, Walter (2014) Don't Until the Final Verb Wait: Reinforcement Learning for Simultaneous. En: Proceedings of the 2014 Conference on Empirical Methods in Natural. pp. 1342-1352 : Association for Computational Linguistics; Disponible en: http://dx.doi.org/10.3115/v1/D14-1140; https://aclanthology.org/D14-1140/. Disponible en: 10.3115/v1/D14-1140.
dc.source.bibliographicCitationKalashnikov, Dmitry; Irpan, Alex; Pastor, Peter; Ibarz, Julian; Herzog, Alexander; Jang, Eric; Quillen, Deirdre; Holly, Ethan; Kalakrishnan, Mrinal; Vanhoucke, Vincent; Levine, Sergey (2018) QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/1806.10293.
dc.source.bibliographicCitationChoi, Eunsol; Hewlett, Daniel; Lacoste, Alexandre; Polosukhin, Illia; Uszkoreit, Jakob; Berant, Jonathan (2017) Hierarchical Question Answering for Long Documents. En: arXiv [cs.CL]. Disponible en: http://arxiv.org/abs/1611.01839.
dc.source.bibliographicCitationDevidze, Rati (2025) Reward Design for Reinforcement Learning Agents. En: arXiv [cs.LG]. Disponible en: http://arxiv.org/abs/2503.21949.
dc.source.instnameinstname:Universidad del Rosario
dc.source.reponamereponame:Repositorio Institucional EdocURspa
dc.subjectProcesamiento de lenguaje natural
dc.subjectTeoría de representación de discursos
dc.subjectAprendizaje por refuerzo
dc.subjectRazonamiento automático
dc.subjectInferencia lógica
dc.subjectRepresentación formal del lenguaje
dc.subject.keywordNatural language processing
dc.subject.keywordDiscourse representation theory
dc.subject.keywordReinforcement learning
dc.subject.keywordAutomatic reasoning
dc.subject.keywordLogic inference
dc.subject.keywordFormal representation of language
dc.titleIntegración del aprendizaje por refuerzo en parsers semánticos para la deducción lógica en lenguaje natural
dc.title.TranslatedTitleIntegration of reinforcement learning in semantic parsers for logical deduction in natural language
dc.typemasterThesis
dc.type.hasVersioninfo:eu-repo/semantics/acceptedVersion
dc.type.spaTesis de maestría
local.department.reportEscuela de Ciencias e Ingeniería
local.regionesBogotá
Archivos
Bloque original
Mostrando1 - 2 de 2
Miniatura
Nombre:
References_aprendizaje_por_refuerzo_en_parsers_semánticos_Gomez_Vargas_Camilo_Andres.ris
Tamaño:
22.62 KB
Formato:
Descripción:
Descargar
Miniatura
Nombre:
Aprendizaje_por_refuerzo_en_parsers_semánticos_Gomez_Vargas_Camilo_Andres.pdf
Tamaño:
4.69 MB
Formato:
Adobe Portable Document Format
Descripción:
Descargar
Colecciones
Maestría en Matemáticas Aplicadas y Ciencias de la Computación