|
Acadêmico(a): LUIZ CÉSAR COPPI JUNIOR |
Título: BRAND FEELING: UM SISTEMA PARA ANALISAR O SENTIMENTO DOS USUÁRIOS EM RELAÇÃO A UMA MARCA |
|
Abstract: |
This work presents the development of a system to analyze the sentiment of Twitter’s users about a particular brand. This system provides an interface where the users can inform a brand and visualize the information related to the corresponding sentiment. When a brand is informed to the system, begins the process of data searching and analysis, starting with the data mining of the advanced search of Twitter. These data are then filtered through techniques to isolate the text of each post, and the application of the Stanford Log-linear Part-Of-Speech (POS) Tagger library, to define the grammatical classes of each term. Afterwards, the MIT Java WordNet Interface (JWI) library along with WordNet is used in order to reduce each term to its radical, as well as the SentiWordNet to verify the value of each term's sentiment. During the development of this work, a category filter to the search was added, aiming to solve an existing limitation on Twitter and to increase the number of terms in each search. As a result, the system was able to analyze a large amount of data coming from Twitter at a satisfactory execution time to the end user. This work has successfully achieved the defined aims, as well as was able to adapt to the limitations encountered throughout the development. This work contribute on the sentiment analysis area throught an application that highlights the potential of this area to the general public, as well as, providing an accessible tool for the regular user. |
|
|