Named entity recognition on emergency response system

• Main Author: Png, Jun Sheng
• Other Authors: Chng Eng Siong
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/148058
• Institution: Nanyang Technological University

Precise and accurate information about a situation is vital in an emergency call so as to ensure that the correct dispatched team arrives at the correct location in a timely manner. However, miscommunication between the dispatcher and the caller may occur which would result in unfortunate events that could have been prevented [1][2]. It is therefore crucial that we uncover solutions to reduce the aforementioned mistakes in order to prevent such mishaps from happening as well as to reduce response time in the Emergency Response System(ERS). This report will discuss the application of using Named Entity Recognition (NER), an information extraction technique on the call between the caller and dispatcher, to improve the efficiency of the ERS by obtaining critical pieces of information and correctly identifying it, thereby reducing or eliminating misunderstandings between the caller and dispatcher. A set of ERS logs will be generated based on grammar rules by using OpenFST and Thrax compiler. The grammar is derived from police calls from videos posted by the community. The logs data set will be trained using the Bi-LSTM-CRF. The final product will be a web demo which allows the user to switch freely between different NER and highlight the correct name entities.