Diploma and Master Theses (authored and supervised):
"Comparing Social Media Topics of Interest Associated with Places According to Userīs Origin";
Supervisor: G. Gartner, F. Porras Bernárdez;
Department für Geodäsie und Geoinformation, FB Kartographie,
final examination: 09-26-2019.
The main theme of the current master thesis is to compare Social Media Topics of Interest associated with Points of Interest according to userīs origin. It is implemented in the two basic phases. The first phase includes the extraction and visualization of the Topics of Interest from Wikipedia Articles, Flickr and Twitter. The second phase deals with the determination of the similarity between Flickr, Twitter and Wikipedia Articles.
The analysis of Wikipedia articles involves the selection of prominent Point of Interests of the city of Vienna and the collection of Wikipedia Articles in German and English language for them. The extraction of the Topics of Interest based on an online tool of creating wordclouds. The next step is to obtain the top ten Topics of Interest or Terms per Point of Interest in both versions (English and German). The visualization of the Topics of Interest or Terms takes place by adding the georeferenced wordclouds to the map. Tables, bar charts, and pie charts depict the most frequent Topics of Interest or Terms.
The analysis of the Geosocial Media data divided into two groups. The first analysis refers to the data derived from Twitter and the second refers to the data derived from Flickr. The analysis of Twitter data contains the clustering of the data by using HDBSCAN algorithm, the extraction of the number of clusters that are related to the Points of Interest and the aggregation of tweets within "related" clusters. The final step of the procedure is to obtain the top ten Topics of Interest or Terms for the related to clusters in English and German language. The most frequent terms depicted with the assistance of tables and bar charts. The analysis of the Flickr data contains the clustering of the data by using HDBSCAN algorithm, the extraction of the number of clusters that are related to the Points of Interest and the aggregation of posts within the "related" clusters. The last step of the process is to obtain the top ten Topics of Interest or Terms for the related clusters. Tables and bar charts are the tools that help us to depict the most frequent terms.
The similarity is "measured" by calculating the frequency of appearance of the Topics of Interest or Terms. The comparison of the percentages of the top ten Terms or Topics of Interest between Twitter, Wikipedia and Flicker is the key to define the similarity. In the current thesis the comparison based into two groups. The first group is Twitter-Wikipedia and the second group is Flicker-Twitter.
Created from the Publication Database of the Vienna University of Technology.