Documenti full-text disponibili:
Abstract
What is corporate reputation? How can it be measured? These two questions have been widely discussed by academics, without coming to a shared definition or evaluation methodology.
Each research gave its own corporate reputation definition, but all the studies agree on one point: corporate reputation is the result of the relationship between a company and its stakeholders. On the stakeholder’s opinion rely most of the corporate rep- utation measurement techniques that have been proposed during the past years, techniques that were criticized.
In this work, we have investigated if corporate reputation can be evaluated from social media data
We focused on the Volkswagen scandal and the buzz it created within the Twitter social network. VW’s scandal was chosen because its widely covered evolution through time and its broad effects on VW’s financial performance.
In order to fulfill the research goal, tweets about VW (from 28/8/15 - to 6/6/16) were collected.
This vast dataset was firstly analyzed and not VW’s related elements were re- moved. The remaining part of it was then classified in two main groups: tweets about VW, but not related to the scandal, and the ones that specifically referred to VW’s wrongdoing. Once the two sets were obtained, each of their elements were evaluated with a sentiment analysis software and after the opinion extraction was calculated the daily aggregated sentiment through a custom-built process, defined to adapt to the Twitter domain. This aggregation produced two different daily sentiment score: the general public opinion about VW and the judgement about the scandal.
The work led to excellent results in the not-relevant elements removal phase and the classification one, but the opinion aggregation did not produce significant outcomes. This final results should not be considered as a research drawback, instead they represent a starting point for further analysis on the opinion creation process.
Abstract
What is corporate reputation? How can it be measured? These two questions have been widely discussed by academics, without coming to a shared definition or evaluation methodology.
Each research gave its own corporate reputation definition, but all the studies agree on one point: corporate reputation is the result of the relationship between a company and its stakeholders. On the stakeholder’s opinion rely most of the corporate rep- utation measurement techniques that have been proposed during the past years, techniques that were criticized.
In this work, we have investigated if corporate reputation can be evaluated from social media data
We focused on the Volkswagen scandal and the buzz it created within the Twitter social network. VW’s scandal was chosen because its widely covered evolution through time and its broad effects on VW’s financial performance.
In order to fulfill the research goal, tweets about VW (from 28/8/15 - to 6/6/16) were collected.
This vast dataset was firstly analyzed and not VW’s related elements were re- moved. The remaining part of it was then classified in two main groups: tweets about VW, but not related to the scandal, and the ones that specifically referred to VW’s wrongdoing. Once the two sets were obtained, each of their elements were evaluated with a sentiment analysis software and after the opinion extraction was calculated the daily aggregated sentiment through a custom-built process, defined to adapt to the Twitter domain. This aggregation produced two different daily sentiment score: the general public opinion about VW and the judgement about the scandal.
The work led to excellent results in the not-relevant elements removal phase and the classification one, but the opinion aggregation did not produce significant outcomes. This final results should not be considered as a research drawback, instead they represent a starting point for further analysis on the opinion creation process.
Tipologia del documento
Tesi di laurea
(Laurea)
Autore della tesi
Barbato, Massimo-Maria
Relatore della tesi
Scuola
Corso di studio
Ordinamento Cds
DM270
Parole chiave
Sentiment analysis,Natural language processing,Text classification
Data di discussione della Tesi
13 Dicembre 2016
URI
Altri metadati
Tipologia del documento
Tesi di laurea
(NON SPECIFICATO)
Autore della tesi
Barbato, Massimo-Maria
Relatore della tesi
Scuola
Corso di studio
Ordinamento Cds
DM270
Parole chiave
Sentiment analysis,Natural language processing,Text classification
Data di discussione della Tesi
13 Dicembre 2016
URI
Statistica sui download
Gestione del documento: