Twitter opinion mining and visualization of Ecuadorian Goverment's decisions
- Authors
- Saquicela Zhunia, Lina Priscila
- Format
- MasterThesis
- Status
- publishedVersion
- Description
Social media impact on the present time is undeniable. People communication dynamics are extremely different from those that were available a decade or fifteen years ago. These new age interactions have allowed people expressing their ideas or opinions in an extended way. On the other hand, such interactions have encouraged social actors to take into consideration these opinions in order to accomplish different goals. Thus, the analysis of data generated by people has become a popular task that is required for different applications. One application can be found at political or government spheres. Specifically, sentiment analysis applications have been implemented in order to evaluate the political scenario before elections or to improve e-government initiatives. Following such approach, the present project introduces the opinion mining of selected Ecuadorian Government?s decisions data generated on Twitter, and the presentation of such results through a Web application. This way, a hybrid sentiment analysis method has been implemented and utilised to classify by its polarity the tweets corresponding to three decisions taken by the Ecuadorian government, which have caused a certain degree of discussion on Twitter. Manual classifications of a portion of the tweet sets have been carried out as well as the utilisation of an online sentiment analysis API for classifying all the tweets and comparing the obtained results. All these implementations and the respective development process allowed evidencing the challenge and limitations of this type of solutions. Although the hybrid method implementation outperformed the sentiment analysis API on some datasets, it was not good enough for all the datasets. In addition the method?s results are clearly far from satisfactory. However the obtained solution can be considered a good starting point that open the doors to future improvements in order to achieve a solid tool.
El presente estudio introduce la miner?a de datos sobre opiniones generadas en Twitter respecto a decisiones tomadas por el Gobierno Ecuatoriano, as? como la presentaci?n de los correspondientes resultados a trav?s de una aplicaci?n web. De esta manera, un m?todo de an?lisis de sentimiento h?brido fue implementado y utilizado para clasificar, seg?n su polaridad, "tweets" correspondientes a tres decisiones que generaron cierto grado de discusi?n en Twitter. Adem?s, una clasificaci?n manual de una porci?n de opiniones fue llevada a cabo as? como la utilizaci?n de una herramienta en l?nea de an?lisis de sentimiento para clasificar todos los tweets y comparar los resultados obtenidos. Se describe tambi?n los procesos de desarrollo e implementaci?n, a trav?s de lo cual se pudo evidenciar los retos y limitaciones de este tipo de soluciones. La soluci?n propuesta fue evaluada en base al c?lculo de medidas de precisi?n para cada m?todo utilizado. los resultados obtenidos muestran que la soluci?n ha sido parcialmente satisfactoria pero se considera como un buen punto de partida para la aplicaci?n de futuras mejoras que permiten consolidar una herramienta s?lida.
- Publication Year
- 2014
- Language
- eng
- Topic
- BIG DATA
BUSINESS INTELLIGENCE
DATA MINING
WEB ENGINEERING
PROGRAMMING
DATA WAREHOUSING
- Repository
- Repositorio SENESCYT
- Rights
- openAccess
- License