Publication: Minería de procesos aplicada al estudio de wikis
Loading...
Official URL
Full text at PDC
Publication Date
2019
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
La creación colaborativa de conocimiento siempre ha sido uno de los pilares de internet desde la web 2.0. A priori, los intentos de crear contenido mediante la acción colectiva de diferentes individuos sin coordinación ni lucración personal debería ser inútil pues se produce un fenómeno conocido como la tragedia de los comunes. La tragedia de los comunes es una situación donde un sistema compuesto por usuarios que actúan independientemente para lograr su beneficio personal tienden a tener comportamientos contrarios al interés común. Sin embargo, se ha demostrado que esto no tiene porque ser cierto: la colaboración desinteresada y sin organización entre los diferentes usuarios ha hecho posible la existencia de comunidades cuyo único propósito es la difusión del conocimiento: las wikis.
Las wikis, compuestas por diferentes artículos, están siendo ampliamente estudiadas. Sin embargo, los procesos que determinan la evolución de su contenido e inherentes a su propio funcionamiento y aquelllos seguidos por los propios usuarios en su actividad no son del todo conocidos. En este proyecto se propone y aplica una serie de técnicas conocidas como (i) minería de procesos para descubrir y analizar estos procesos existentes en la labor de la escritura colaborativa tanto a nivel artículo como a nivel usuario así como (ii) técnicas de minería social para visualizar las estructuras de colaboración existentes entre los propios usuarios. Para esto se hará uso de la Wikipedia Española como referencia.
Con el objetivo de realizar este estudio se hará uso de los historiales de revisión con el que cuenta cada artículo de Wikipedia y una taxonomía de intenciones semánticas tras cada revisión compuesta de 13 categorías como contra vandalismo, refactorización o elaboración. Haciendo uso de un conjunto de datos dotado 5684 revisiones y sus intenciones semánticas se desarrolla un modelo predictivo que alcanza un valor de F1 micro de 0.64. Con este modelo y dichos historiales de revisiones se genera un corpus compuesto de diferentes artículos destacados y las intenciones tras cada una de sus revisiones. Con las revisiones en combinación con sus intenciones y el uso de minería social y de procesos se observa la estructura colaborativa de los usuarios, los procesos seguidos por los artículos así como los procesos seguidos por los propios usuarios en sus sesiones de edición.
Los resultados muestran que, aunque no existe un proceso unificado en la evolución de los artículos, se puede ver como las diferentes maneras de trabajar de los editores en etapas tempranas del artículo tiene influencia en el desarrollo del mismo. Además, los procesos seguidos por los propios usuarios siguen patrones que permiten clasificarlos dentro de una taxonomía de roles de trabajo, verificando los hallazgos obtenidos en otros estudios. Por último, aunque generalmente no existe colaboración explícita entre los usuarios, se observan colaboraciones organizadas en momentos puntuales.
Collaborative writing has always been one of the pillars of the internet since the web 2.0. Usually, the attempts to create content collaboratively with individuals without organization or benefit are useless, resulting in a phenomenon called the tragedy of the commons. The tragedy of the commons, is a situation where a system composed of independent users that pursue their own goals behave against the common good. However, the selfless collaboration between different, unorganized users made possible the existence of communities whose only purpose is the diffusion of knowledge: the wikis. The wikis, formed by a corpus of different articles, are currently under extensive study. Despite that, the processes that determine the evolution of its content and the processes followed by the users in its activity are not totally known. In this project, a series of techniques are proposed and applied: (i) process mining to discover and analyze the existent processes in the task of collaborative writing from an article and user point of view, and (ii) social mining to visualize the collaboration structures among the different users. The object of the study, will be the Spanish Wikipedia. Every article in Wikipedia has a record of all the editions made to it. Those records along with a taxonomy of 13 semantic intention behind each revision (e.g re-factoring, elaboration or counter vandalism) will be used to meet the goals of this study. With a data set of 5684 revisions and its semantic intentions, a predictive model with a F1 micro of 0.64 is created. Combining this model with the records of revisions of some featured articles generates the different semantic intentions behind each revision. These predicted semantic intentions and its revisions constitute the input of the techniques of social mining and process mining. Such techniques allows the observation of the processes followed by the article, the processes followed by their users in their activity and the collaborative structure between users. Results show that, even though there is no such thing as a unified process in the evolution of the articles, the different behaviours of the users in the initial stages of an article have an influence in its development. Furthermore, the processes followed by the users in their activity follow patterns that verify the findings of previous studies in this topic. Generally, there is no organized collaboration among the users. However, the results imply that sometimes the collaboration between users is explicitly organized.
Collaborative writing has always been one of the pillars of the internet since the web 2.0. Usually, the attempts to create content collaboratively with individuals without organization or benefit are useless, resulting in a phenomenon called the tragedy of the commons. The tragedy of the commons, is a situation where a system composed of independent users that pursue their own goals behave against the common good. However, the selfless collaboration between different, unorganized users made possible the existence of communities whose only purpose is the diffusion of knowledge: the wikis. The wikis, formed by a corpus of different articles, are currently under extensive study. Despite that, the processes that determine the evolution of its content and the processes followed by the users in its activity are not totally known. In this project, a series of techniques are proposed and applied: (i) process mining to discover and analyze the existent processes in the task of collaborative writing from an article and user point of view, and (ii) social mining to visualize the collaboration structures among the different users. The object of the study, will be the Spanish Wikipedia. Every article in Wikipedia has a record of all the editions made to it. Those records along with a taxonomy of 13 semantic intention behind each revision (e.g re-factoring, elaboration or counter vandalism) will be used to meet the goals of this study. With a data set of 5684 revisions and its semantic intentions, a predictive model with a F1 micro of 0.64 is created. Combining this model with the records of revisions of some featured articles generates the different semantic intentions behind each revision. These predicted semantic intentions and its revisions constitute the input of the techniques of social mining and process mining. Such techniques allows the observation of the processes followed by the article, the processes followed by their users in their activity and the collaborative structure between users. Results show that, even though there is no such thing as a unified process in the evolution of the articles, the different behaviours of the users in the initial stages of an article have an influence in its development. Furthermore, the processes followed by the users in their activity follow patterns that verify the findings of previous studies in this topic. Generally, there is no organized collaboration among the users. However, the results imply that sometimes the collaboration between users is explicitly organized.
Description
Trabajo de Fin de Grado, Universidad Complutense, Facultad de Informática, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2018/2019