Authors: Rajiv Paudel*, Michigan State University
Topics: Spatial Analysis & Modeling, Qualitative Research
Keywords: Agent-based Models, Qualitative Data, Natural Language Processing, Unsupervised Data Extraction
Session Type: Poster
Start / End Time: 1:20 PM / 3:00 PM
Room: Napoleon Foyer/Common St. Corridor, Sheraton, 3rd Floor
Presentation File: No File Uploaded
Qualitative data is extremely valuable for agent-based models (ABMs) development as it helps in better understanding the system and developing justifiable models. However, translating qualitative information to the model is not straightforward, and literature lacks structures for efficient data extraction and model development. Existing approaches generally involve manual data coding for supervised information extraction. We live in a data inundated world where more than 2.5 billion terabytes of online data, mostly consisting of unstructured data, is produced every day. Modeling community will benefit immensely if they can use such vast datasets that are readily available at their fingertips. However, conventional supervised data extraction methods are time and resource consuming and are not advisable for large and unstructured qualitative data. This study outlines a loosely supervised technique that can extract information for ABM development from large unstructured qualitative datasets with minimum manual work. Using Natural Language Processing (NLP) technique, the study uses semi-structured field interviews to develop ABM to analyze household vulnerability to food insecurity. The method can be extended to extract information for ABM development from large textual data effortlessly.