Computing Similarity between Tweets in Text, Time and Space
(master thesis)
Computing similarities between Tweets is of crucial importance for a number of application areas like disaster management, urban planning, or fight against crime and terrorism. However in contrast to most previous natural language processing (NLP) approaches, which focused purely on textual content, the approach addressed in this master’s thesis implicitly considers the temporal and spatial dimensions, which carry vital information. This thesis builds on existing research, which developed an interdisciplinary method for emotion classification that combines linguistic, temporal, and spatial information into a single similarity metric and subsequently applies a graph-based semi-supervised learning approach to labels all tweets with an emotion class. The main goal of this thesis is to improve the current algorithm:
- by increasing the efficiency through the development of a new tweet labelling algorithm, and
- by validating the definition of linguistic, spatial and temporal similarity parameters.
Contact: Willi Mann, Nikolaus Augsten, Bernd Resch
In cooperation with:
