Sentiment Structures in Street Harassment Stories
Advisor Information
Parvathi Chundi
Location
Dr. C.C. and Mabel L. Criss Library
Presentation Type
Poster
Start Date
7-3-2014 1:00 PM
End Date
7-3-2014 4:00 PM
Abstract
Street harassment is a pervasive problem that typically targets women and LGBTQ community. Hollaback! is an international movement aimed at stopping street harassment. Hollaback! servers collect street harassment stories from victims around the globe to share, gather statistics, and create awareness. In this research, we present a simple method aimed at identifying attack severity in Hollaback! stories. Since the Hollaback! dataset encompasses many different levels of attack severity--catcalling, groping, stalking, and assault--and all these levels of attack severity are generally negative, our method analyzes the sentiments and weights at the sentence level to identify harassment severity. Sentence level sentiments are used to construct the sentiment structure of a story which is used to identify subsets of severity amongst Hollaback! stories. The proposed methods are applied to a Hollaback! data set containing around 1900 stories (written in English) posted from different cities in the United States, and cities around the world. Our experimental results illustrate the power of the proposed method.
Sentiment Structures in Street Harassment Stories
Dr. C.C. and Mabel L. Criss Library
Street harassment is a pervasive problem that typically targets women and LGBTQ community. Hollaback! is an international movement aimed at stopping street harassment. Hollaback! servers collect street harassment stories from victims around the globe to share, gather statistics, and create awareness. In this research, we present a simple method aimed at identifying attack severity in Hollaback! stories. Since the Hollaback! dataset encompasses many different levels of attack severity--catcalling, groping, stalking, and assault--and all these levels of attack severity are generally negative, our method analyzes the sentiments and weights at the sentence level to identify harassment severity. Sentence level sentiments are used to construct the sentiment structure of a story which is used to identify subsets of severity amongst Hollaback! stories. The proposed methods are applied to a Hollaback! data set containing around 1900 stories (written in English) posted from different cities in the United States, and cities around the world. Our experimental results illustrate the power of the proposed method.