Patricio Santiago García-Montero
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Patricio Santiago García-Montero has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Jorge Luis Zambrano-Martinez
Universidad del Azuay
8 shared publications - Marcos Orellana
University of Azuay
8 shared publications - Alexandra Bermeo
Universidad del Azuay
1 shared publication - Andrés Patiño-León
Universidad del Azuay
1 shared publication - Anthony Xavier Romero Gonzalez
Universidad del Azuay
1 shared publication García- Montero
1 shared publication
Latest Publications
- Open Access
- 20 January 2026
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 32 (1) , 472-483
AI-Driven Hierarchical Taxonomy Generation from Emergency Call Transcripts
This article presents a case study on hierarchical topic modeling for emergency call transcripts from Ecuador's ECU 911 service. We introduce a hybrid methodology that first generates a taxonomy from unlabeled data using BERTopic and agglomerative clustering, and then employs embedding-based similarity for multi-label classification. By leveraging multilingual embeddings (LaBSE) and clustering algorithms (UMAP & HDBSCAN), we identified 23 coherent topics, demonstrating a practical balance between accuracy and operational applicability. The key result is a significant reduction in Hamming Loss and an F1-score of 0.4951, achieved without the need for pre-labeled data. This underscores the method's primary practical significance: offering a scalable, automated solution for emergency management centers to rapidly categorize complex incidents, thereby enhancing situational awareness and resource allocation. The integration of LLaMA 3 for automated label generation further optimized semantic interpretation, highlighting the potential of language models in critical, resource-constrained domains.
- 20 January 2026
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 32 (1) , 472-483
Open AccessAI-Driven Hierarchical Taxonomy Generation from Emergency Call Transcripts
This article presents a case study on hierarchical topic modeling for emergency call transcripts from Ecuador's ECU 911 service. We introduce a hybrid methodology that first generates a taxonomy from unlabeled data using BERTopic and agglomerative clustering, and then employs embedding-based similarity for multi-label classification. By leveraging multilingual embeddings (LaBSE) and clustering algorithms (UMAP & HDBSCAN), we identified 23 coherent topics, demonstrating a practical balance between accuracy and operational applicability. The key result is a significant reduction in Hamming Loss and an F1-score of 0.4951, achieved without the need for pre-labeled data. This underscores the method's primary practical significance: offering a scalable, automated solution for emergency management centers to rapidly categorize complex incidents, thereby enhancing situational awareness and resource allocation. The integration of LLaMA 3 for automated label generation further optimized semantic interpretation, highlighting the potential of language models in critical, resource-constrained domains.
- 17 October 2025
- Published by Springer Nature
- p. 92-105
Boosting Automatic Speech Recognition Performance for Noisy Emergency Calls with Hybrid Processing Techniques
- 17 October 2025
- Published by Springer Nature
- p. 92-105
Boosting Automatic Speech Recognition Performance for Noisy Emergency Calls with Hybrid Processing Techniques
- 17 October 2025
- Published by Springer Nature
- p. 63-77
Enriching Dataset Metadata with LLMs to Unlock Semantic Meaning
- 17 October 2025
- Published by Springer Nature
- p. 63-77
Enriching Dataset Metadata with LLMs to Unlock Semantic Meaning
- Open Access
- 16 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 179759-179775
Legal AI for All: Reducing Perplexity and Boosting Accuracy in Normative Texts With Fine-Tuned LLMs and RAG
This study investigates the potential of open-source Large Language Models (LLMs) to enhance the accessibility and understanding of complex legal information. Legal texts often present linguistic and structural barriers for non-experts, highlighting the need for tools that can support natural language interaction with normative content. We propose a framework that integrates Fine-Tuning and Retrieval-Augmented Generation (RAG) techniques to adapt LLMs for domain-specific legal applications. Using synthetic, validated question-answer datasets, we train and evaluate multiple open-source models and introduce novel benchmarks to assess performance across different difficulty levels. Results show that the fine-tuned models significantly reduce perplexity (up to 10×) and, when combined with RAG, achieve accuracy levels exceeding 90%. To demonstrate adaptability, an additional use case involving travel-related legal norms in an environmentally protected insular region is included. This confirms the scalability of the approach to diverse legal scenarios, supporting its potential for broader adoption in legal information systems.
- 16 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 179759-179775
Open AccessLegal AI for All: Reducing Perplexity and Boosting Accuracy in Normative Texts With Fine-Tuned LLMs and RAG
This study investigates the potential of open-source Large Language Models (LLMs) to enhance the accessibility and understanding of complex legal information. Legal texts often present linguistic and structural barriers for non-experts, highlighting the need for tools that can support natural language interaction with normative content. We propose a framework that integrates Fine-Tuning and Retrieval-Augmented Generation (RAG) techniques to adapt LLMs for domain-specific legal applications. Using synthetic, validated question-answer datasets, we train and evaluate multiple open-source models and introduce novel benchmarks to assess performance across different difficulty levels. Results show that the fine-tuned models significantly reduce perplexity (up to 10×) and, when combined with RAG, achieve accuracy levels exceeding 90%. To demonstrate adaptability, an additional use case involving travel-related legal norms in an environmentally protected insular region is included. This confirms the scalability of the approach to diverse legal scenarios, supporting its potential for broader adoption in legal information systems.
- 10 October 2024
- Published by Springer Nature
- p. 238-252
Incident Alert Priority Levels Classification in Command and Control Centre Using Word Embedding Techniques
- 10 October 2024
- Published by Springer Nature
- p. 238-252
Incident Alert Priority Levels Classification in Command and Control Centre Using Word Embedding Techniques
- 10 October 2024
- Published by Springer Nature
- p. 131-142
Automatic Parking Space Segmentation Using K-Means Clustering and Image Processing Techniques
- 10 October 2024
- Published by Springer Nature
- p. 131-142
Automatic Parking Space Segmentation Using K-Means Clustering and Image Processing Techniques
- 10 October 2024
- Published by Springer Nature
- p. 271-284
Pre-processing of the Text of ECU 911 Emergency Calls
- 10 October 2024
- Published by Springer Nature
- p. 271-284
Pre-processing of the Text of ECU 911 Emergency Calls
- Open Access
- 29 December 2023
- Published by MDPI AG in Big Data and Cognitive Computing
- Vol. 8 (1) , 3
Semantic Similarity of Common Verbal Expressions in Older Adults through a Pre-Trained Model
Health problems in older adults lead to situations where communication with peers, family and caregivers becomes challenging for seniors; therefore, it is necessary to use alternative methods to facilitate communication. In this context, Augmentative and Alternative Communication (AAC) methods are widely used to support this population segment. Moreover, with Artificial Intelligence (AI), and specifically, machine learning algorithms, AAC can be improved. Although there have been several studies in this field, it is interesting to analyze common phrases used by seniors, depending on their context (i.e., slang and everyday expressions typical of their age). This paper proposes a semantic analysis of the common phrases of older adults and their corresponding meanings through Natural Language Processing (NLP) techniques and a pre-trained language model using semantic textual similarity to represent the older adults’ phrases with their corresponding graphic images (pictograms). The results show good scores achieved in the semantic similarity between the phrases of the older adults and the definitions, so the relationship between the phrase and the pictogram has a high degree of probability.
- 29 December 2023
- Published by MDPI AG in Big Data and Cognitive Computing
- Vol. 8 (1) , 3
Open AccessSemantic Similarity of Common Verbal Expressions in Older Adults through a Pre-Trained Model
Health problems in older adults lead to situations where communication with peers, family and caregivers becomes challenging for seniors; therefore, it is necessary to use alternative methods to facilitate communication. In this context, Augmentative and Alternative Communication (AAC) methods are widely used to support this population segment. Moreover, with Artificial Intelligence (AI), and specifically, machine learning algorithms, AAC can be improved. Although there have been several studies in this field, it is interesting to analyze common phrases used by seniors, depending on their context (i.e., slang and everyday expressions typical of their age). This paper proposes a semantic analysis of the common phrases of older adults and their corresponding meanings through Natural Language Processing (NLP) techniques and a pre-trained language model using semantic textual similarity to represent the older adults’ phrases with their corresponding graphic images (pictograms). The results show good scores achieved in the semantic similarity between the phrases of the older adults and the definitions, so the relationship between the phrase and the pictogram has a high degree of probability.
- Open Access
- 15 October 2023
- Published by Escuela Superior Politecnica del Litoral in Revista Tecnológica - ESPOL
- Vol. 35 (2) , 192-205
Síntesis Textual de Evaluación para Acoso y Ciberacoso
En los últimos años, el acoso y el ciberacoso son problemas que han aumentado vertiginosamente afectando escuelas, colegios y universidades. Debido a los avances en las tecnologías de la información, cualquier persona está expuesta a ser atacada; por esta razón, es necesario crear soluciones a través de técnicas adecuadas que ayuden a prevenir el acoso y ciberacoso. En consecuencia, en este artículo se propone crear una síntesis textual a partir de datos de encuestas que permita desarrollar modelos para clasificar o predecir tanto a víctimas como agresores de acoso y ciberacoso. Para ello, se utilizaron técnicas de minería de datos, árboles de decisión y técnicas de agrupación, dando como resultado una síntesis textual. Esto permitió la creación y evaluación de un modelo de aprendizaje supervisado y otro modelo con técnicas de agrupamiento, aplicadas a los datos de las encuestas realizadas a estudiantes universitarios. Los resultados demostraron la importancia de la síntesis textual para la generación de modelos de clasificación o predicción de víctimas y agresores del acoso y ciberacoso, con una exactitud mayor al 75%, siendo el modelo de agrupamiento con mejor rendimiento.
- 15 October 2023
- Published by Escuela Superior Politecnica del Litoral in Revista Tecnológica - ESPOL
- Vol. 35 (2) , 192-205
Open AccessSíntesis Textual de Evaluación para Acoso y Ciberacoso
En los últimos años, el acoso y el ciberacoso son problemas que han aumentado vertiginosamente afectando escuelas, colegios y universidades. Debido a los avances en las tecnologías de la información, cualquier persona está expuesta a ser atacada; por esta razón, es necesario crear soluciones a través de técnicas adecuadas que ayuden a prevenir el acoso y ciberacoso. En consecuencia, en este artículo se propone crear una síntesis textual a partir de datos de encuestas que permita desarrollar modelos para clasificar o predecir tanto a víctimas como agresores de acoso y ciberacoso. Para ello, se utilizaron técnicas de minería de datos, árboles de decisión y técnicas de agrupación, dando como resultado una síntesis textual. Esto permitió la creación y evaluación de un modelo de aprendizaje supervisado y otro modelo con técnicas de agrupamiento, aplicadas a los datos de las encuestas realizadas a estudiantes universitarios. Los resultados demostraron la importancia de la síntesis textual para la generación de modelos de clasificación o predicción de víctimas y agresores del acoso y ciberacoso, con una exactitud mayor al 75%, siendo el modelo de agrupamiento con mejor rendimiento.