Ponencias (Lenguajes y Sistemas Informáticos)

URI permanente para esta colecciónhttps://hdl.handle.net/11441/11394

Examinar

Mostrando 1 - 20 de 1332

Counterfactual Simulation for estimating Performance Loss in PV Systems: A Machine Learning approach
(Springer Science and Business Media Deutschland GmbH, 2026) Sánchez López, José Enrique; Luna Romera, José María; Mateos García, Daniel; Galán Sales, Francisco Javier; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia, Innovación y Universidades (MICIU). España; European Union
Performance loss in photovoltaic systems (PV) is caused by multiple factors, such as soiling, panel degradation, and operational maintenance. Traditional estimation approaches rely on statistical methods to isolate degradation signals from environmental conditions. This study proposes a novel counterfactual simulation framework based on machine learning (ML) to address this challenge. The core idea is to train a separate regression model for each year using SCADA data from that year and then use it to simulate how the system would have performed under the environmental conditions of other years. This version of counterfactual prediction reconstructs past operational states and isolates changes in energy output that can be attributed to long-term system degradation rather than fluctuating weather. Two visual diagnostics support this analysis: a degradation envelope capturing forecast dispersion over time and a training-bias comparison that highlights how models trained on different years diverge in their forecasts. The methodology was applied to four datasets from the Desert Knowledge Australia Solar Center (DKASC), resulting in performance loss between 0.75% and 1.75% per year. These results align with expectations, as only environmental variability was decoupled while maintenance-related losses remained unmodeled.
NVS-HO: A Benchmark for Novel View Synthesis of Handheld Objects
(Springer Science and Business Media Deutschland GmbH, 2027) Ali, Musawar; Carranza García, Manuel; Fioraio, Nicola; Salti, Samuele; Di Stefano, Luigi; Lenguajes y Sistemas Informáticos
We propose NVS-HO, the first benchmark designed for novel view synthesis of handheld objects in real-world environments using only RGB inputs. Each object is recorded in two complementary RGB sequences: (1) a handheld sequence, where the object is manipulated in front of a static camera, and (2) a board sequence, where the object is fixed on a ChArUco board to provide accurate camera poses via marker detection. The goal of NVS-HO is to learn a NVS model that captures the full appearance of an object from (1), whereas (2) provides the groundtruth images used for evaluation. To establish baselines, we consider both a classical SfM pipeline and a state-of-the-art pre-trained feed-forw ard neural network (VGGT) as pose estimators, and train NVS models based on NeRF and Gaussian Splatting. Our experiments reveal significant performance gaps in current methods under unconstrained handheld conditions, highlighting the need for more robust approaches. NVS-HO thus offers a challenging real-world benchmark to drive progress in RGB-based novel view synthesis of handheld objects.
Task Scheduling Optimization on Enterprise Application Integration Platforms Based on the Meta-heuristic Particle Swarm Optimization
(ASSOC COMPUTING MACHINERY, 2017) Sellaro, Daniela F.; Frantz, Rafael Z.; Hernández Salmerón, Inmaculada Concepción; Roos-Frantz, Fabricia; Sawicki, Sandro; Lenguajes y Sistemas Informáticos
Companies seek technological alternatives that provide competitiveness for their business processes. Among these alternatives, there are integration platforms that allow you to connect applications to your software ecosystems. These ecosystems are often composed of local applications and cloud computing services, such as SaaS and PaaS, and still, interact with social media. Integration platforms are specialized software that allows you to design, execute and monitor integration solutions, which connect functionality and data from different applications. Integration platforms typically provide a specific domain language, development toolkit, runtime engine, and monitoring tool. The efficiency of the engine in scheduling and performing integration tasks has a direct impact on the performance of a solution and this is one of the challenges faced by integration platforms. Our literature review has identified that integration engines adopt task scheduling algorithms based on the textit First-In-First-Out discipline, which may be inefficient. Therefore, it is appropriate to seek a task scheduling algorithm that optimizes engine performance, providing a positive impact on the performance of the integration solution in different scenarios. This article proposes an algorithm for task scheduling based on the meta-heuristic optimization technique, which assigns the tasks to the computational resources, considering the waiting time in the queue of ready tasks and the computational complexity of Each task in order to optimize the performance of the integration solution.
Cloud Configuration Modelling: a Literature Review from an Application Integration Deployment Perspective
(Elsevier BV, 2015) Hernández Salmerón, Inmaculada Concepción; Sawicki, Sandro; Carneiro Roos, Fabricia; Frantz, Rafael Z.; Lenguajes y Sistemas Informáticos
Enterprise Application Integration has played an important role in providing methodologies, techniques and tools to develop integration solutions, aiming at reusing current applications and supporting the new demands that arise from the evolution of business processes in companies. Cloud-computing is part of a new reality in which companies have at their disposal a high capacity IT infrastructure at a low-cost, in which integration solutions can be deployed and run. The charging model adopted by cloud-computing providers is based on the amount of computing resources consumed by clients. Such demand of resources can be computed either from the implemented integration solution, or from the conceptual model that describes it. It is desirable that cloud-computing providers supply detailed conceptual models describing the variability of services and restrictions between them. However, this is not the case and providers do not supply the conceptual models of their services. The conceptual model of services is the basis to develop a process and provide supporting tools for the decision-making on the deployment of integration solutions to the cloud. In this paper, we review the literature on cloud configuration modelling, and compare current proposals based on a comparison framework that we have developed.
Hacia la construcción de gemelos digitales de procesos no estructurados de organizaciones
(Sistedes, 2023) Bravo, Alfonso; Peña Siles, Joaquín; del Río Ortega, Adela; Resinas Arias de Reyna, Manuel; Lenguajes y Sistemas Informáticos; Durán Toro, Amador; TIC205: Ingeniería del Software Aplicada
Un gemelo digital de una organización se puede entender como un modelo dinámico de ésta que utiliza datos operacionales de la organización junto a otra información contextual para entender cómo funciona la organización, predecir su comportamiento y actuar en caso de que éste se aleje de los objetivos deseados. En los últimos años se han desarrollado propuestas que hacen uso de técnicas de minería de procesos y simulación para construir gemelos digitales en base a los procesos de la organización. Sin embargo, estas propuestas no abordan el reto que supone considerar la gran cantidad de trabajo no estructurado que se realiza en las organizaciones, por lo que sólo modelan una visión parcial de la misma. En este artículo, partimos de la premisa de que ese trabajo no estructurado queda reflejado en plataformas colaborativas y presentamos nuestra visión para construir gemelos digitales que aborda esta carencia. Además, identificamos cuatro retos para desarrollarlos y esbozamos formas de abordarlos.
A Hybrid Reliability Metric for SLA Predictive Monitoring
(ASSOC Computing Machinery, 2019) Comuzzi, Marco; Márquez Chamorro, Alfonso Eduardo; Resinas Arias de Reyna, Manuel; Lenguajes y Sistemas Informáticos; Ministerio de Economia, Industria y Competitividad (MINECO). España; TIC205: Ingeniería del Software Aplicada
Modern SLA management includes SLA prediction based on data collected during service operations. Besides overall accuracy of a prediction model, decision makers should be able to measure the reliability of individual predictions before taking important decisions, such as whether to renegotiate an SLA. Measures of reliability of individual predictions provided by machine learning techniques tend to depend strictly on the technique chosen and to neglect the features of the system generating the data used to learn a model, i.e., the service provisioning landscape in this case. In this paper, we consider business process-aware service provisioning and we define a hybrid measure of reliability of an individual SLA prediction for classification models, which accounts for both the reliability of the chosen prediction technique, if available, and features capturing the variability of the service provisioning scenario. The metric is evaluated empirically using SLAs and process event logs of a real world case.
Using Large Language Models to Develop Requirements Elicitation Skills
(ACM, 2025) Lojo, Nelson; González, Rafael; Philip, Rohan; Parejo Maestre, José Antonio; Durán Toro, Amador; Fox, Armando; Fernández Montes, Pablo; Lenguajes y Sistemas Informáticos; TIC205: Ingeniería del Software Aplicada
Requirements Elicitation (RE) is a crucial software engineering skill that involves interviewing a client and then devising a software design based on the interview results. We propose conditioning a large language model to play the role of the client during a chat-based interview. We evaluate our approach in a study (𝑛 = 120) using both a qualitative survey and quantitative observations about participants’ work. Our positive findings suggest a new way to practice critical RE skills in a scalable and realistic manner without the overhead of arranging live interviews.
HORIZON: A Classification and Comparison Framework for Pricing-Driven Feature Toggling
(Springer, 2026) García Fernández, Alejandro; Parejo Maestre, José Antonio; Ruíz Cortés, Antonio; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España
The rise of the Software as a Service (SaaS) paradigm has popularized subscription-based models, enabling user-specific customization but complicating the enforcement of pricing constraints across the codebase. Feature toggles are a common tool for managing dynamic behavior, but applying them to pricing-driven environments introduces new challenges that, among others, the vision of pricing-driven development and operation aims to address. While some studies have pointed out the lack of support in current industrial tools and the limited scope of academic proposals, the specific improvements needed to fully realize this vision remain unclear. To fill this gap, we present HORIZON, a framework for classifying and comparing feature toggling tools in pricing-driven contexts. By applying it to existing literature, the framework was able to highlight key strengths, weaknesses, and research opportunities, guiding the development of more robust solutions for pricing-aware SaaS.
iSubscription: Bridging the Gap Between Contracts and Runtime Access Control in SaaS
(Springer, 2026) García Fernández, Alejandro; Parejo Maestre, José Antonio; Ruíz Cortés, Antonio; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España
Software as a Service (SaaS) usually provides users with the ability to choose the features and usage limits they need –i.e. a configuration. Once a configuration is selected, it becomes a subscription. Despite enhancing value for customers, this model challenges providers, who must enforce that users operate within subscriptions that may change at any time –either by user choice or provider updates. This makes runtime subscription enforcement a self-adaptation challenge. A promising strategy to address this challenge is pricing-driven feature toggling, currently implemented only by Pricing4SaaS, to the best of our knowledge. This solution relies on two data sources: iPricings –machine-oriented representations of pricings; and subscription states –usage levels’ values of usage-limited features. However, Pricing4SaaS delegates subscription state management to the managed service, forcing providers to update subscription logic whenever a new usage limit is added/removed, which increases coupling and hinders scalability. This paper introduces three main contributions to address these limitations. First, we show that us-age limits tend to increase as pricings evolve, reinforcing the need to simplify subscription management. Second, we introduce iSubscription, a machine-oriented model of subscription states compatible with iPricings. Third, we introduce SPACE, an independent service for pricing-driven self-adaptation that overcomes all state-of-the-art’s limitations.
Racing the Market: An Industry Support Analysis for Pricing-Driven DevOps in SaaS
(Springer, 2025) García Fernández, Alejandro; Parejo Maestre, José Antonio; Cavero, Francisco Javier; Ruíz Cortés, Antonio; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España
The SaaS paradigm has popularized the usage of pricings, allowing providers to offer customers a wide range of subscription possibilities. This creates a vast configuration space for users, enabling them to choose the features and support guarantees that best suit their needs. Regardless of the reasons why changes in these pricings are made, the frequency of changes within the elements of pricings continues to increase. Therefore, for those responsible for the development and operation of SaaS, it would be ideal to minimize the time required to transfer changes in SaaS pricing to the software and underlying infrastructure, without compromising the quality and reliability. This work explores the support offered by the industry for this need. By modeling over 150 pricings from 30 different SaaS over six years, we reveal that the configuration space grows exponentially with the number of add-ons and linearly with the number of plans. We also evaluate 21 different feature toggling solutions, finding that feature toggling, particularly permission toggles, is a promising technique for enabling rapid adaptation to pricing changes. Our results suggest that developing automated solutions with minimal human intervention could effectively reduce the time-to-market for SaaS updates driven by pricing changes, especially with the adoption of a standard for serializing pricings.
Towards Effective SaaS Pricing Design: A Case Study of CCSIM
(Springer, 2026) García Fernández, Alejandro; Laso, Sergio; Parejo Maestre, José Antonio; Berrocal, Javier; Ruíz Cortés, Antonio; Murillo, Juan Manuel; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España
The widespread adoption of Software as a Service (SaaS) has led providers to offer diverse pricing to meet varying customer needs. However, there is currently no methodology for eliciting and designing these structures. In this paper, we present our experience in defining a pricing for CCSIM, an advanced simulator designed for deploying virtualized, highly customizable computing continuum architectures that initially lacked one. As a result, CCSIM improved its transparency for customers –allowing them to better understand the offered features and their associated costs– and increased provider profitability. Building on this practical experience and towards our vision of pricings as software artifacts, we devise a first approach towards a process that we term “SaaS pricification” –a method for defining consistent, easily manageable, and profitable pricings. In short, this work introduces SaaS pricification as a step towards integrating pricings into SaaS products lifecycle, offering an initial approach derived from the practical experience of pricifying CCSIM.
Automated Analysis of Pricings in SaaS-Based Information Systems
(Springer, 2025) García Fernández, Alejandro; Parejo Maestre, José Antonio; Trinidad, Pablo; Ruíz Cortés, Antonio; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España
Software as a Service (SaaS) pricing models, encompassing features, usage limits, plans, and add-ons, have grown exponentially in complexity, evolving from offering tens to thousands of configuration options. This rapid expansion poses significant challenges for the development and operation of SaaS-based Information Systems (IS), as manual management of such configurations becomes time-consuming, error-prone, and ultimately unsustainable. The emerging paradigm of Pricing-driven DevOps aims to address these issues by automating pricing management tasks, such as transforming human-oriented pricings into machine-oriented (iPricing) or finding the optimal subscription that matches the requirements of a certain user, ultimately reducing human intervention. This paper advances the field by proposing seven analysis operations that partially or fully support these pricing management tasks, thus serving as a foundation for defining new, more specialized operations. To achieve this, we mapped iPricings into Constraint Satisfaction Optimization Problems (CSOP), an approach successfully used in similar domains, enabling us to implement and apply these operations to uncover latent, yet non-trivial insights from complex pricing models. The proposed approach has been implemented in a reference framework using MiniZinc, and tested with over 150 pricing models, identifying errors in 35 pricings of the benchmark. Results demonstrate its effectiveness in identifying errors and its potential to streamline Pricing-driven DevOps.
Towards a Framework for Multi-Bot Collaboration
(IEEE Computer SOC, 2025) Mimbrero, Alberto; Parejo Maestre, José Antonio; Fernández Montes, Pablo; Romero Arjona, Miguel; Segura Rueda, Sergio; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España; TIC205: Ingeniería del Software Aplicada
Software bots are rapidly proliferating, assisting developers with tasks such as coding, testing, modeling, and documentation. However, most research has focused on the use of isolated bots, overlooking a critical aspect to unlock their full potential: multi-bot collaboration. In this work-in-progress paper, we introduce the first prototype of Botica, a framework leveraging asynchronous APIs, messaging, and containerization to facilitate the development, deployment, and interaction of bots. A pilot study using Botica for REST API testing illustrates its effectiveness and the potential advantages of bot collaboration.
ASTRAL: Automated Safety Testing of Large Language Models
(IEEE Computer SOC, 2025) Ugarte, Miriam; Valle, Pablo; Parejo Maestre, José Antonio; Segura Rueda, Sergio; Arrieta, Aitor; Lenguajes y Sistemas Informáticos; TIC205: Ingeniería del Software Aplicada
Large Language Models (LLMs) have recently gained significant attention due to their ability to understand and generate sophisticated human-like content. However, ensuring their safety is paramount as they might provide harmful and unsafe responses. Existing LLM testing frameworks address various safety-related concerns (e.g., drugs, terrorism, animal abuse) but often face challenges due to unbalanced and obsolete datasets. In this paper, we present ASTRAL, a tool that automates the generation and execution of test cases (i.e., prompts) for testing the safety of LLMs. First, we introduce a novel black-box coverage criterion to generate balanced and diverse unsafe test inputs across a diverse set of safety categories as well as linguistic writing characteristics (i.e., different style and persuasive writing techniques). Second, we propose an LLM-based approach that leverages Retrieval Augmented Generation (RAG), few-shot prompting strategies and web browsing to generate up-to-date test inputs. Lastly, similar to current LLM test automation techniques, we leverage LLMs as test oracles to distinguish between safe and unsafe test outputs, allowing a fully automated testing approach. We conduct an extensive evaluation on well-known LLMs, revealing the following key findings: i) GPT3.5 outperforms other LLMs when acting as the test oracle, accurately detecting unsafe responses, and even surpassing more recent LLMs (e.g., GPT-4), as well as LLMs that are specifically tailored to detect unsafe LLM outputs (e.g., LlamaGuard); ii) the results confirm that our approach can uncover nearly twice as many unsafe LLM behaviors with the same number of test inputs compared to currently used static datasets; and iii) our black-box coverage criterion combined with web browsing can effectively guide the LLM on generating up-to-date unsafe test inputs, significantly increasing the number of unsafe LLM behaviors.
AI-Driven Fairness Testing of Large Language Models: A Preliminary Study
(IEEE, 2025) Romero Arjona, Miguel; Parejo Maestre, José Antonio; Alonso Valenzuela, Juan Carlos; Sánchez Jerez, Ana Belén; Arrieta, Aitor; Segura Rueda, Sergio; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España; TIC205: Ingeniería del Software Aplicada
Fairness is a fundamental principle of trustworthy Artificial Intelligence systems, yet it remains difficult to assess and enforce. Existing fairness testing methods depend heavily on manual evaluation and predefined templates or datasets, which are resource-intensive and limit their scalability and applicability. In this work-in-progress paper, we establish the foundation for a fully automated approach to fairness testing in large language models (LLMs) based on two main ideas. First, we propose applying metamorphic testing to identify bias by analysing how model responses change when modifications are made to input prompts. Second, we propose using LLMs for both test case generation and output evaluation, leveraging their capability to generate diverse inputs and classify outputs effectively. A pilot study shows the potential of our approach to uncover bias in three widely used LLMs: Gemma, Llama3, and Mistral. However, the study also reveals challenges to be addressed to ensure broader applicability, providing a basis for future research in this critical field.
SATORI: Static Test Oracle Generation for REST APIs
(IEEE, 2025) Alonso Valenzuela, Juan Carlos; Martín López, Alberto; Segura Rueda, Sergio; Bavota, Gabriele; Ruiz Cortés, Antonio; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; Agencia Estatal de Investigación. España; TIC205: Ingeniería del Software Aplicada
REST API test case generation tools are evolving rapidly, with growing capabilities for the automated generation of complex tests. However, despite their strengths in test data generation, these tools are constrained by the types of test oracles they support, often limited to crashes, regressions, and noncompliance with API specifications or design standards. This paper introduces SATORI (Static API Test ORacle Inference), a black-box approach for generating test oracles for REST APIs by analyzing their OpenAPI Specification. SATORI uses large language models to infer the expected behavior of an API by analyzing the properties of the response fields of its operations, such as their name and descriptions. To foster its adoption, we extended the PostmanAssertify tool to automatically convert the test oracles reported by SATORI into executable assertions. Evaluation results on 17 operations from 12 industrial APIs show that SATORI can automatically generate up to hundreds of valid test oracles per operation. SATORI achieved an F1-score of 74.3%, outperforming the state-of-the-art dynamic approach AGORA+ (69.3%)—which requires executing the API—when generating comparable oracle types. Moreover, our findings show that static and dynamic oracle inference methods are complementary: together, SATORI and AGORA+ found 90% of the oracles in our annotated ground-truth dataset. Notably, SATORI uncovered 18 bugs in popular APIs (Amadeus Hotel, Deutschebahn, FDIC, GitLab, Marvel, OMDb and Vimeo) leading to documentation updates by the API maintainers.
Interruptibility during Scientific Research Collaborations: The Effect of Pressure, Proximity and Quiet Time
(ACM, 2024-11-13) Beerepoot, Iris; Río Ortega, Adela del; Resinas Arias de Reyna, Manuel; Reijers, Hajo A.; Lenguajes y Sistemas Informáticos; Farzan, Rosta; López, Claudia; Ministerio de Ciencia e Innovación (MICIN). España; European Union (UE); TIC205: Ingeniería del Software Aplicada
Work disruptions may derail scientific research projects by interrupting the flow of work and information transfer. In this study, we explore the interruptibility of scientific researchers and its correlation with contextual factors such as deadlines, outside-work hours, and location. We report in-progress findings from an autoethnographic study building on an analysis of interaction logs, as well as location and calendar data. This analysis provides a unique perspective on research projects that run for several months, revealing situations in which we allow ourselves to be interrupted. Our preliminary results provide metrics that highlight periods of high interruptibility and present initial results of a correlation analysis between interruptibility and work context. Our findings reveal that as deadlines approach, the urgency leads to longer uninterrupted work periods and the work is prioritized over other projects. Evening and weekend work is also more conducive to sustained work periods, whereas summer work and work done in close proximity to others is generally more fragmented.
UVL.js: Experiences on using UVL in the JavaScript Ecosystem
(ACM, 2025) Lamas, Víctor; Limaylla Lunarejo, María Isabel; Luaces, Miguel R.; Romero Organvidez, David; Galindo Duarte, José Ángel; Benavides Cuevas, David Felipe; Lenguajes y Sistemas Informáticos; Acher, Mathieu; Alves Pereira, Juliana; Ministerio de Ciencia e Innovación (MICIN). España; European Union (UE); TIC276: Diverso Lab - International Computing
The Universal Variability Language (UVL) was developed as a community-driven effort to create a simple yet extensible language for feature modeling, promoting tool interoperability within the software product line community. Although UVL is supported by several tools like FeatureIDE, Flamapy, and Pure::variants, it currently lacks direct support for web environments. To address this, we introduce a JavaScript-based UVL parser built with the ANTLR framework. This parser makes UVL models accessible directly within browser-based environments, eliminating the need for extra installations and enhancing UVL’s usability for web-based tools. Furthermore, the parser can be used in back-end environments with JavaScript runtime environments such as Node.js. The parser has been successfully tested with more than 1,000 UVL models available on UVLHub and supports various UVL language levels and conversion strategies. We demonstrate its integration through two use cases: UVLHub, a public repository for UVL models developed using open science principles, and an application lifecycle management tool for software product lines. This JavaScript UVL parser is the first of its kind, unlocking new possibilities for web and JavaScript applications to take advantage of the advancements in UVL technology.
AquaIA: Implementing Controlled Deficit Irrigation by Means of Feature Modelling Techniques
(ACM, 2025) Benitez, Francisco Sebastian; Galindo Duarte, José Ángel; Corell González, Mireia; Domínguez Mayo, Francisco José; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia, Innovación y Universidades (MICIU). España; TIC276: Diverso Lab - International Computing
Andalusia’s agriculture is heavily dependent on irrigated crops. Currently it is under threat due to a severe and prolonged drought. To address this crisis, innovative water-saving strategies such as Controlled Deficit Irrigation (CDI) are being explored. CDI reduces water input during non-critical growth stages of crops, yet its successful application requires precise calibration based on crop type and soil characteristics. This paper presents AquaIA , a variability aware platform that leverages Software Product Line Engineering (SPLE) techniques to tailor CDI strategies. By modelling irrigation schema configurations using the Universal Variability Language (UVL), AquaIA enables the creation of optimized irrigation systems through an accessible configurator. Experimental results show water consumption reductions of up to 50% for tomatoes and 30% for lettuces, demonstrating the platform’s potential for sustainable agriculture in water-scarce regions.
Context-Aware AI Agents for Clinical Dialogue Assistance through Large Language Models
(Springer, 2026) Naranjo Pozas, Rodrigo; Doblado Mendoza, Pablo; Vega Márquez, Belén; Lenguajes y Sistemas Informáticos; Ministerio de Ciencia e Innovación (MICIN). España; TIC134: Sistemas Informáticos
This research investigates the efficacy of Artificial Intelligence Agents in processing and responding to personalized, private contextual information. We studied how to implement a system designed to augment an open-source Large Language Model (LLM), such as Llama, Claude, and Gemma, with domain-specific knowledge bases. This augmentation is intended to facilitate the generation of contextually coherent and accurate responses to user queries. The system was developed and tested using different versions of a clinical conversation transcripts dataset between patients and medical professionals, enabling specialized knowledge integration. The architectural framework, built upon LangChain, FAISS, Ollama, and Gradio, demonstrates a simple, modular, scalable, and extensible design. This work helped us to take our first steps into the development of robust AI agents capable of leveraging external knowledge for enhanced conversational intelligence in specialized domains.

Examinar

Envíos recientes