Acquiring Stored Or Real Time Satellite Data Via Natural Language Query

With the advent of Sensor Web, the satellite data acquired by sensor systems could be shared among users immediately. Our research has led to an implementation of natural language queries such that users without particular knowledge of satellite imagery can describe easily for what they need. We use a rules-based method to retrieve named entities, with the help of a knowledge base and uses existing Sensor Web services for acquiring stored or real time satellite data. We use rule-based methods to align time, location and domain task entities in natural language queries with Sensor Web services with standard times, geographical coordinates, and satellite attributes. To evaluate our system, we wrote a series of natural language queries in the domains of surveying and mapping, forestry, agriculture, and disaster response. Our queries and satellite data retrieved by the queries were corrected by a group of experts to create a gold standard. Using their remarks as correct, we scored our system results using precision and recall metrics standard for information retrieval. The results of our experiment demonstrate that the proposed method is promising for assisting in Earth observation applications.