In Partial Fulfillment of the Requirements for the Degree of Master of Science
will defend his thesis
Emotion-Weighted Density Estimation and its Application to Emotion Mapping
Twitter tweets have become a valuable source for emotion analysis. In this research, we use tweets for emotion mapping. The inputs for our approach are geo-tweets which are annotated with emotion scores that range from -1 to +1. Traditional density estimation techniques consider only the spatial dimension of data points ignoring non-spatial information, such as emotion scores. In this research we investigate novel emotion-weighted density estimation approaches that differ from the traditional non-parametric density estimation by additionally considering a non-spatial variable of interest in its influence function. For measuring the density at a location, we take an input set of objects that are characterized by a location and an emotion-value. From this input, a 2-dimensional continuous density function that takes negative and positive values is generated. The density function is then used to create maps capturing the variation of happiness in an observation area. We develop an efficient grid-based algorithm for computing the density at a query point. We also analyze the influence of the parameters, such as bandwidth on the shape of the generated density function. Moreover, this thesis develops visualization techniques for emotion-weighted density functions. Straightforward 3D visualizations as well as grid-based 2D displays are developed for this purpose. Analyzing change in spatial data is critical for many applications; to support this need, this research also develops animation techniques to facilitate density change visualization for spatial datasets. Finally, we evaluate the developed techniques and frameworks for producing tweet-based happiness maps for the State of New York.
Date: Monday, April 22, 2019
Time: 4:30 PM
Place: PGH 550
Advisor: Dr. Christoph F. Eick
Faculty, students, and the general public are invited.