A Semantic Enhanced Model for Searching in Spatial Web Portals

Extended Abstract Spatial Web Portal (SWP, Yang et al 2007) is used by the Earth science community in Earth science data sharing and exchanging. The SWP facilitates a large amount of geospatial resources, including text files, raw and post-processed data, and various geospatial web services. However, the popular utilizations also stand out problems of how to find needed data from a variety of geospatial resources and how to visualize the data from multi-perspectives. Currently, most search engines in SWPs are based on keyword matching or pure ontology based reasoning, which can not effectively 'understand' the meaning of user's queries, especially when a user has limited Earth science knowledge. For example, after California fire, people may ask more about " What's the air pollution caused by Southern California Fire in 2007? " (We define it as query Q1), but usually they do not know what pollutant may cause air pollution, or what factor is used in Air Quality (AQ) community to measure how bad the air pollution is. The above information is domain knowledge that is hidden behind the query and can not be inferred directly by keyword based search engines. Thus how to provide a 'specialized' answer through 'unspecialized' queries becomes an urgent challenge, which is also known as " Intelligent Question Answering " problem. To solve the problem, this extended abstract reports our efforts on utilizing semantic web techniques (Berners-Lee, Hendler and Lassila, 2001) to enhance traditional search engines by: 1) building a semantic-based information model, which is a network of concepts with explicit relationships for implementing knowledge reasoning. With this model, a generic query can be explained and inferred to specific information needed. 2) Combining search tools provided within and cross SWPs to combine multi-layers of heterogeneous data sources together, results will include both web pages for presenting text information and spatial maps generated from remote web services for interactive search. Our model is based on SWEET (Semantic Web for Earth and Environmental Terminology) ontology, where all the concepts are divided into different facets, including phenomena, property, substance, earth realm, to support reductionism (Raskin and Pan, 2005). As the most popular ontology model in Earth Science, SWEET provides an upper-level abstracted expression of the world. Based on formal description logic (DL), it can support terminology reasoning (T-box reasoning). However, only using this model may not be enough for answering query like Q1. First, " Southern California Fire in …