Objectives Remote sensing image demand text is a user-friendly approach for users to acquire images through natural language. Time is a key element when searching for remote sensing images, thus extracting and understanding time description in remote sensing image demand text is critical in-demand semantic analysis. However, a few pieces of research focus on time extraction and semantic computing of remote sensing image demand text, and time extraction and semantic computing are not combined to get full consideration of time semantic analysis.
Methods we collect remote sensing image demand texts from domains of disaster mitigation, agriculture, forestry, surveying and mapping. Then we extract time descriptions from demand texts manually and analyze their composition. By dividing time description into different time elements, a formal representation model is provided and time semantic representation rules as while as time semantic computation rules are proposed. Finally, a time ontology that integrates the time formal representation and semantic computing together is developed, which including four parts: Time semantic elements, time description, time model, and time knowledge. A rule-based time extracting method and a semantic computing method which both based on the time ontology are also given.
Results We carried out experiments on 2 000 remote sensing image demand texts, which were divided into five groups. The baseline was made manually and compared with the results of our methods. We got an average precision of 97.1%. Experiments show that our method has better efficiency for understanding time information in remote sensing data demand text.
Conclusions Integrating time extraction and semantic computing into one ontology providing a seamless comprehension of time description in remote sensing image demand text, which helps to improve the efficiency of obtaining remote sensing images through natural language.
DownLoad: