1. Interpretable classification and summarization of crisis events from microblogs
- Author
-
Nguyen, Thi Huyen and Nguyen, Thi Huyen
- Abstract
The widespread use of social media platforms has created convenient ways to obtain and spread up-to-date information during crisis events such as disasters. Time-critical analysis of crisis-related information helps humanitarian organizations and governmental bodies gain actionable information and plan for aid response. However, situational information is often immersed in a high volume of irrelevant content. Moreover, crisis-related messages also vary greatly in terms of information types, ranging from general situational awareness - such as information about warnings, infrastructure damages, and casualties - to individual needs. Different humanitarian organizations or governmental bodies usually demand information of different types for various tasks such as crisis preparation, resource planning, and aid response. To cope with information overload and efficiently support stakeholders in crisis situations, it is necessary to (a) classify data posted during crisis events into fine-grained humanitarian categories, (b) summarize the situational data in near real-time. In this thesis, we tackle the aforementioned problems and propose novel methods for the classification and summarization of user-generated posts from microblogs. Previous studies have introduced various machine learning techniques to assist humanitarian or governmental bodies, but they primarily focused on model performance. Unlike those works, we develop interpretable machine-learning models which can provide explanations of model decisions. Generally, we focus on three methods for reducing information overload in crisis situations: (i) post classification, (ii) post summarization, (iii) interpretable models for post classification and summarization. We evaluate our methods using posts from the microblogging platform Twitter, so-called tweets. First, we expand publicly available labeled datasets with rationale annotations. Each tweet is annotated with a class label and rationales, which are short snippe
- Published
- 2024