| dc.description.abstract | The use of Geographic Information Systems (GIS) has expanded across various areas and has become even more important with the growth of big data in recent decades. However, its accessibility can be limited, because geo-analytical tools are spread out across multiple software programs and also scattered within a single software environment. To address this issue, one possible solution is to develop a geo-analytical Question-Answering (QA) system. Unlike traditional QA systems that work with simple query task, a geoGIS QA system requires a more complex transformation task. In order to achieve this, the answers provided by a GeoQA system should be in a workflow format, while the user’s questions are expressed in natural language. In this context, our research focuses on bridging the gap between workflows and natural language by introducing the concept of subquestions, which describe an underlying task, along with their corresponding sub-workflows. For this study, we generated data in the form of (sub-)questions, (sub-)workflows, and parse trees. These generated data were then subjected to analysis using two methods. The first method involves measuring the similarity between sub-questions and between parse trees to understand the extent to which GIS tools impact changes in sentences. The second method is based on word subtraction, which aims to identify specific question fragments associated with GIS tools. The methodology has demonstrated its ability to find patterns in natural language sentences and connect them to GIS tools, which are part of workflow structures. These patterns can potentially contribute to the development of a geoQA system, taking us a step closer towards its realization. | |
