When: Wednesday, February 27, 2019
Where: PGH 563
Time: 11:00 AM

A Novel Data Analysis Framework to Understand US Emotions in Space and Time

Romita Banerjee, PhD Student

My research centers on the development of a novel spatio-temporal clustering framework, named Makalu, which assesses the happiness and well-being of a population over time. As Twitter is one of the most popular social media platforms used by millions of users, we use it as a knowledge source for emotion analysis. The input for the framework is the location and time of where and when the tweets were posted and an emotion assessment score in the range [-1, +1], with +1 representing a very positive emotion and -1 representing a very negative emotion. This input is then segmented into batches with each batch containing tweets that occur in a specific time interval; for example, daily batches. Next, by generalizing the existing kernel density estimation techniques, each batch is transformed into a continuous function that takes positive and negative values. Finally, contouring algorithms are used to find contiguous regions with highly positive and highly negative emotions for each of the batch. The framework also proposes animation techniques to facilitate spatio-temporal data storytelling based on the obtained spatio-temporal data analysis results.

Romita Banerjee is a PhD student in the UH Data Analysis and Intelligence Systems Lab, Department of Computer Science, University of Houston. Prior to joining UH, she received her master’s in computer science from Florida State University, Tallahassee. Her research focuses on spatio-temporal clustering and data mining

Study and Mitigation of Platform Related UWB Ranging Errors

Nour Smaoui, PhD Student

Ultra Wideband technology has been presented as a precise and accurate solution for wireless ranging. However, with low-cost off-the-shelf UWB devices, noise and errors can be very challenging. To solve this problem, a comprehensive characterization of these errors is required for these devices to work correctly. In this work, we cover the error characterization in different environments and configurations, propose a correction method for Line-of-sight scenarios and evaluate its performance. The results show that after correction, the system presents errors of a maximum mean of 8 cm.

Nour Smaoui received his undergraduate degree in Computer Science from The National School of Computer Sciences, Tunisia in 2014. Subsequently, he worked at Sagemcom Software and Technologies for a year, developing software for set-top boxes. Nour is presently a PhD student and Teaching Assistant in Department of Computer Science at University of Houston. His research interests include Wireless Networks and Wireless Localization. He has taught or has assisted in teaching a diverse range of courses.

Emphasis Selection by Learning from Label Distributions

Reza (Amirreza) Shirani , PhD Student

Emphasizing words in a text is considered as a common and effective way to captivate audiences attention. Emphasis is the strengthening of words in a text in a different style from the rest of the text, to highlight them. Emphasis can clarify or even change the meaning of a sentence and it can be done by using different colors, shapes, italics, boldface and different graphic marks. Some existing layout algorithms and graphic design applications like Adobe Spark emphasize words by providing automatic text layouts for text segments. However, these algorithms know nothing about the meaning of the text, and so they lay out the text based on non-semantic attributes like word length. In this study, we investigate models that aim at suggesting to the user which text segment(s) can be emphasized based on semantic of the text. Word emphasis patterns are person- and domain-specific. To tackle the multi-label nature of this task we deploy a label distribution learning model to find better ways to model the label ambiguity in emphasis selection problem.

Reza (Amirreza) Shirani received his undergraduate degree in Computer Engineering from Shahid Beheshti University, Iran in 2013 and his Master’s degree in the same major from University of Isfahan, Iran in 2016. Reza is currently a Ph.D. student and Research Assistant at RiTUAL LAB, Department of Computer Science at University of Houston. His research interests include Natural Language Processing, Computational Linguistics, Applied Machine Learning, Deep Learning.