Publicación: Detección y alerta de noticias falsas en procesos electorales con una metodología basada en el estructuralismo narrativo
Fecha
2023-09-01
Autores
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Lenguajes y Sistemas Informáticos
Resumen
La desinformación no es un problema nuevo, pero ha adquirido dimensiones preocupantes en la última década con la polarización de la política y la expansión de las redes sociales. También ha demostrado ser un fenómeno especialmente grave cuando actúa en procesos electorales, desvirtuando la voluntad popular y arrebatando a los ciudadanos el derecho a estar informados cuando acuden a las urnas; el derecho a elegir libremente, en definitiva. Los algoritmos de detección de noticias falsas y las agencias de verificación intentan mitigar este problema, pero en la mayoría de los casos solo consiguen contrarrestarlo cuando el bulo ya se ha extendido demasiado. El presente trabajo propone un modelo de detección y alerta temprana de noticias falsas, basado en el estructuralismo narrativo iniciado por Vladimir Propp. Este antropólogo y lingüista ruso analizó más de un centenar de cuentos y leyendas rusos para encontrar las estructuras subyacentes que los hacían similares pese a contar historias aparentemente distintas. A diferencia de otros sistemas, que analizan el contenido, la forma o el contexto de las noticias falsas, esta propuesta estudia la estructura que sujeta los mensajes informativos, esos cuentos y leyendas que se repiten en los procesos electorales de casi todo el mundo. El modelo propuesto ha conseguido una mejora de más del 6% con respecto al modelo preentrenado base. Se propone también un sistema que jerarquiza los bulos de acuerdo con su peligrosidad, para que se priorice su comprobación antes de que se extiendan demasiado. Este trabajo parte de una colección de casi 3.400 tuiteos recopilados durante las Elecciones generales de Brasil 2022, que incluye 77 mensajes desinformativos anotados manualmente.
Disinformation is not a new issue, but it has taken on alarming dimensions over the past decade, fueled by political polarization and the proliferation of social media platforms. It has proven to be an especially grave phenomenon when it influences electoral processes, distorting the popular will and depriving citizens of their right to be informed when casting their votes; essentially, their right to make a free choice. Algorithms designed to detect fake news and fact-checking agencies attempt to mitigate this issue, yet in most cases they only manage to counteract it after the falsehood has already spread widely. The present study proposes a model for early detection and alert of fake news, grounded in the narrative structuralism initiated by Vladimir Propp. This Russian anthropologist and linguist analyzed over a hundred Russian tales and legends to uncover underlying structures that made them similar despite narrating apparently different stories. Unlike other systems that analyze the content, form, or context of fake news, this proposal examines the structure that underpins news stories—those tales and legends that recur in electoral processes worldwide. The proposed model has achieved an improvement of more than 6% compared to the baseline pre-trained model. Additionally, a system is proposed that ranks falsehoods according to their level of potential damage, so that their verification is prioritized before they spread too extensively. This study is based on a collection of nearly 3,400 tweets gathered during the General Elections in Brazil in 2022, which includes 77 manually annotated disinformation messages.
Disinformation is not a new issue, but it has taken on alarming dimensions over the past decade, fueled by political polarization and the proliferation of social media platforms. It has proven to be an especially grave phenomenon when it influences electoral processes, distorting the popular will and depriving citizens of their right to be informed when casting their votes; essentially, their right to make a free choice. Algorithms designed to detect fake news and fact-checking agencies attempt to mitigate this issue, yet in most cases they only manage to counteract it after the falsehood has already spread widely. The present study proposes a model for early detection and alert of fake news, grounded in the narrative structuralism initiated by Vladimir Propp. This Russian anthropologist and linguist analyzed over a hundred Russian tales and legends to uncover underlying structures that made them similar despite narrating apparently different stories. Unlike other systems that analyze the content, form, or context of fake news, this proposal examines the structure that underpins news stories—those tales and legends that recur in electoral processes worldwide. The proposed model has achieved an improvement of more than 6% compared to the baseline pre-trained model. Additionally, a system is proposed that ranks falsehoods according to their level of potential damage, so that their verification is prioritized before they spread too extensively. This study is based on a collection of nearly 3,400 tweets gathered during the General Elections in Brazil in 2022, which includes 77 manually annotated disinformation messages.
Descripción
Categorías UNESCO
Palabras clave
Citación
Centro
Facultades y escuelas::E.T.S. de Ingeniería Informática
Departamento
Lenguajes y Sistemas Informáticos