Análisis de sentimiento en eventos con contrincantes
Loading...
Official URL
Full text at PDC
Publication date
2017
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
En este trabajo consideramos el problema de la caracterización de mensajes y usuarios en las redes sociales, en este caso el servicio de microblogging Twitter. Para ello hemos descargado 13,3 millones de tweets utlizando como contexto las elecciones generales estadounidenses que tuvieron lugar en el mes de noviembre de 2016. A partir de los tweets descargados construimos un conjunto coherente mediante un proceso de limpieza de datos. Una vez construido, analizamos el conjunto utilizando la técnica del análisis de sentimiento para asignar de forma automática una etiqueta descriptiva a cada tweet que indica si el autor del tweet apoya a alguno de los candidatos o, por lo contrario, se opone activamente a alguno de los candidatos.
Una vez compilados los datos obtenemos una serie de resultados, especialmente donde analizamos el comportamiento de los usuarios que apoyan a un candidato con respecto al oponente. Encontramos que los seguidores del candidato republicano Donald Trump fueron más activos en Twitter y más beligerantes contra la candidata Hillary Clinton que viceversa.
Finalmente, comparamos nuestro conjunto de datos con un estudio similar en Twitter.
This paper considers the problem of categorizing text and users in the context of social networking, in this case the microblogging service Twitter. To to this we downloaded over 13 million tweets in the days leading up to the 2016 U.S presidential elections. Using the tweets we downloaded we used a process known as data cleansing to build a coherent dataset. Once the dataset was built, we used sentiment analysis to automatically assign a label indicating whether the tweet's author supports one of the candidates, or on the contrary, actively opposes them. Once every tweet was labeled we compiled a series of results, the most interesting of which being analyzing the behavior of users who support a candidate while opposing the other. We found that supporters of the republican candidate Donald Trump were more active on Twitter and more belligerent against the democratic candidate Hillary Clinton than vice versa. Finally, we compare our dataset to the dataset of a similar study.
This paper considers the problem of categorizing text and users in the context of social networking, in this case the microblogging service Twitter. To to this we downloaded over 13 million tweets in the days leading up to the 2016 U.S presidential elections. Using the tweets we downloaded we used a process known as data cleansing to build a coherent dataset. Once the dataset was built, we used sentiment analysis to automatically assign a label indicating whether the tweet's author supports one of the candidates, or on the contrary, actively opposes them. Once every tweet was labeled we compiled a series of results, the most interesting of which being analyzing the behavior of users who support a candidate while opposing the other. We found that supporters of the republican candidate Donald Trump were more active on Twitter and more belligerent against the democratic candidate Hillary Clinton than vice versa. Finally, we compare our dataset to the dataset of a similar study.
Description
Trabajo de Fin de Grado en Ingeniería Informática (Universidad Complutense, Facultad de Informática, curso 2016/2017)












