Autor:
Molino-Peña, Elena

Cargando...
Foto de perfil

E-mails conocidos

mmolino@us.es

Fecha de nacimiento

Proyectos de investigación

Unidades organizativas

Puesto de trabajo

Apellidos

Molino-Peña

Nombre de pila

Elena

Nombre

Nombres alternativos

Molino Peña, Elena

Afiliaciones conocidas

Universidad de Sevilla, Spain

Páginas web conocidas

Página completa del ítem
Notificar un error en este autor

Resultados de la búsqueda

Mostrando 1 - 3 de 3
  • Artículo
    Operaciones de Análisis sobre los Términos de Uso en Customer Agreements
    Molino-Peña, Elena; García, José María; Ruiz-Cortés, Antonio. Actas de las XVII Jornadas de Ciencia e Ingeniería de Servicios (JCIS 2022), 2022-09-05.
    La gestión y contratación de servicios cloud es una actividad cada vez más esencial en muchas organizaciones. Independientemente del tipo (IaaS, PaaS o SaaS) un servicio cloud dispone de un acuerdo de cliente o Customer Agreement (CA) que recoge los términos y condiciones generales sobre la prestación del servicio. Debido a la magnitud y complejidad de los CA, su comprensión suele resultar una tarea difícil para el cliente y el proveedor, además de propensa a errores. Una posible aproximación para mitigar este problema es automatizar la interpretación de CA. Con el fin de crear en el futuro una herramienta software que lo soporte, en este primer estudio nos centramos en identificar y clasificar algunas preguntas claves que podría realizarse el proveedor y el cliente en diferentes contextos, al crear, valorar o ejecutar el acuerdo. Este catálogo de preguntas se ha validado mediante el análisis manual del CA de GitHub.
  • Artículo
    Limitations of current techniques to detect Obligations and Rights in SLA
    Molino-Peña, Elena; García, José María; Ruiz-Cortés, Antonio. Actas de las XVIII Jornadas de Ciencia e Ingeniería de Servicios (JCIS 2023), 2023-09-12.
    The need to know and understand the obligations and rights contained in service agreements (also known as Customer Agreement) has generated interest in the industry, and there are numerous projects and tools available for the automatic detection and interpretation of obligations and rights. This is especially beneficial for the customer, as it allows them to automatically know the commitments and risks associated with the use of cloud services, as well as for the provider, who can detect potential liability gaps in the agreement. However, existing tools are only able to extract and partially identify this information. In this study, three limitations have been identified in the patterns proposed by some of the recent techniques to extract information from service level agreements (SLAs). In particular, situations in which some obligation or right is not detected, causes for which they can be misclassified and scenarios in which the actor that performs the action is not detected. In addition, a possible solution to these obstacles is proposed.
  • Artículo
    Analysing Customer Agreements with LLMs: Is GPT-4 better than me?
    Molino-Peña, Elena; García, José María; Ruiz-Cortés, Antonio. Actas de las XIX Jornadas de Ciencia e Ingeniería de Servicios (JCIS 2024), 2024-06-17.
    The growing need to provide automatic answers to questions about legal documents has generated an enormous interest from both academia and industry. Although there are tools for the automatic detection and reasoning of contract information, automatic analysis is still a developing area of research. This becomes even more relevant with advances in Artificial Intelligence (AI) and the boom in cloud services, highlighting the need to explore the performance of Large Language Models (LLMs) in analysing Customer Agreements (CAs). Based on a previous article in which we proposed a catalogue of 37 analysis operations on terms of use, this study compares the answers obtained manually with those generated by GPT-4. The results obtained are promising, showing that in most cases LLMs are able to respond like a human agent, but there are some situations where they may differ. These mistakes can be caused by both the human and the LLM, and are often due to the lack or saturation of information provided, or misinterpretation of the text.