View Item 
        •   Utrecht University Student Theses Repository Home
        • UU Theses Repository
        • Theses
        • View Item
        •   Utrecht University Student Theses Repository Home
        • UU Theses Repository
        • Theses
        • View Item
        JavaScript is disabled for your browser. Some features of this site may not work without it.

        Browse

        All of UU Student Theses RepositoryBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

        The Online Presence of Liveability: Using sentiment derived from Twitter messages as an indicator of liveability.

        Thumbnail
        View/Open
        GIMA Thesis Mattijs de Haas 5722195 31052021.pdf (2.507Mb)
        Publication date
        2021
        Author
        Haas, M.J.M. de
        Metadata
        Show full item record
        Summary
        High liveability is associated with benefits like socio-economic equity, inclusive social systems, and improved mental and physical health. Assessing and afterwards effectively improving liveability has been a goal of local, national, and international governments and initiatives. The ambiguity of the concept of liveability complicates its assessment. This study examined the use of residents’ sentiment as an indicator of liveability in the Netherlands and explored ways to collect residents’ sentiment from location-based social networks. In this study, eight methods for deriving sentiment from Twitter messages were compared. The best performing method, based on Youden’s J and percentage of tweets classified, was used to classify the of index dataset of Twitter messages from 2019 located within the Netherlands to sentiment. A total of 1 375 203 tweets was classified, and the resulting sentiment was grouped to the 380 municipalities of the Netherlands. The sentiment scores were compared to the expected liveability patterns, and to an existing liveability index from the Netherlands: the Leefbaarometer 2.0, version of 2018. A Naïve Bayes’ classifier performed the best in the performance assessment, with a Youden’s J of 0,46. The mean sentiment per municipality showed no similarities to the expected liveability patterns on a national scale, but the expected large regional differences in the peripheral regions were present in the sentiment scores. The mean sentiment and the existing liveability index showed a weak positive relation in the cursory visual analysis, but no statistical relation was found. In conclusion, the sentiment derived from twitter messages in this study does not significantly represent liveability. Exploring additional sentiment classification methods and using more training data could improve the quality of the sentiment analysis, further solidifying the conclusive strength of research on sentiment and liveability. Another valuable path to follow in future research includes dissecting liveability and analysing the relation of sentiment to its different aspects.
        URI
        https://studenttheses.uu.nl/handle/20.500.12932/1138
        Collections
        • Theses
        Utrecht university logo