Geospatial Semantic Web Blog

The W3C Geospatial Incubator Group recently published two reports on geospatial vocabularies and ontologies. The announcement came through the W3C Semantic Web News:

The final reports of the W3C Geospatial Incubator Group has just been published: Geospatial Vocabulary and Geospatial Ontologies. The first document also includes a reference to a GeoOWL ontology that relies on other, existing vocabularies (like GeoRSS or GML). The second document gives an overview of some other, existing ontologies in the area.

These two documents provide a good overview of existing vocabularies and ontologies. Use cases of geospatial ontologies are described in the Geospatial Vocabulary document. Based on my reading, the group has reached an agreement on a model for representing geometry objects (points, lines, polygons, etc.) but not yet on a model for expressing geospatial relationships (within, touches, overlaps, etc.).

GeoRSS seems to play an influential role in the described model. This is a good sign. To application developers, ontology construction is a means to an end. RSS and GeoRSS extensions enabled the creation of many useful applications. Building on these standards will encourage a broader community to embrace a Geospatial Wemantic Web.

The ability to represent time and temporal relations is an important aspect of the Semantic Web. W3C recently published a new working draft on Time Ontology in OWL. This work is an updated version of the OWL-Time ontology developed by Jerry Hobbs and Feng Pan.

Before the standardization of the OWL language, Jerry and Feng also worked on a version of the ontology in DAML. Their previous works on time ontologies can be found at Feng’s OWL-Time website.

In the RDF and OWL world, there are at least two different ways to represent time. One way is to use XSD datatypes (e.g., xsd:datetime), and the other is to use OWL-Time. Sometimes people ask me which approach is better for representing time in their applications.

Read the rest of this entry »

The term ontology should be no stranger to anyone who studies the Semantic Web. Even if it’s the first time that you’ve heard of this term, you can answer “what’s ontology?” by googling Wikipedia — “an ontology is a data model that represents a domain and is used to reason about the objects in that domain and the relations between them.”

On the Web, what counts as an ontology? This is a question that Tim Finin and Li Ding have explored in their paper “Untangling ontologies on the Semantic Web“.

In their research, they analyzed a collection of over 1.7 million Semantic Web documents (RDF documents) that were crawled by the Swoogle search engine. Based on the statistic data from their analysis, they inferred the characteristics of ontologies in the present Web and used this knowledge to answer the question: what counts as an ontology?

Read the rest of this entry »

Geospatial ontologies are formal representations of geographical concepts and relations. At the Networking Geospatial Information Technology, Jerry Hobbs of ISI gave an interesting presentation on understanding and developing geospatial ontologies.

His thesis is that in order to develop geospatial semantic web applications, we must first develop a core theory of geospatial and other spatial representation and reasoning. In particular, he discussed ontology development from a natural language perspective — e.g., how do we model queries such as “How long is chile?”, “How large is N. Korea?”, “How far is LA from Washington DC, as the crow flies?”

He also brought up a lot of interesting issues such as scales and half orders of magnitude.

Read the rest of this entry »

Swoogle is a Semantic Web search engine developed by researchers at the Ebiquity Research Group at UMBC. As of June 5th, 2006, the Swoogle crawler has found 1.5 million semantic web documents.

These 1.5M documents comprise about 1M RDF documents, 350K documents with embedded RDF data and 150K documents that look like Semantic Web documents but are currently inaccessible or fail to parse properly. About 3000 new documents are discovered each day. We estimate that of the 1M RDF documents, about 1% (10K) are ontologies, as opposed to data, examples or test files.

Here is how you can discover geospatial ontology using Swoogle.

(1) Find any ontology documents that mention keyword “geo”

• Enter search string “geo“

(2) Find only OWL documents that mention keyword “geo”

• Enter search string “hasFiletype:owl geo“

(3) Find only RDF documents that mention keyword “geo” and “latitude”

• Enter search string “hasFiletype:owl geo latitude“

(4) Find any ontology documents that mention keyword “geometry” but not “algebra”

• Enter search string “geometry NOT algebra“

(5) Find all RSS feeds that contains the term “geo”

• Click on the “document” search option on the top of the search box
• Enter search string “hasFiletype:rss geo“

For more search options, see Swoogle user guide.

After reading this blog entry, a reader asks the following:

The definition [of geospatial semantics] is convincing. However, could you tell the difference between the “Geospatial semantics” and Egenhofer and Mark (1995)s’ “Naive Geography” which defines as : “The body of knowledge that people have about the surrounding geographic world.”

In this post, I will try to answer Andrea’s question.

I believe there are similarities and differences between the studies of Naive Geography and Geospatial Semantics.

Read the rest of this entry »

Prof. Tim Finin posted a video of our first Semantic Web application: ITTalks. It’s a talk announcement system that publishes web contents in both human and machine readable format.

This system was developed under the DARPA DAML program. Part of the existing Web Ontology Language OWL is based on an ontology language (i.e., DAML+OIL) developed under this program. In 2000, semantic web research was mostly focused on how to publish machine readable web contents and annotate web information using shared ontological vocabularies.

If you watch the video and compare the ITTalks system with some recent Semantic Web applications, you see obvious difference in their research focus and feature capabilities.

From my personal perspective, since I have worked on the ITTalks project and still in the Semantic Web field, I think Semantic Web research has evolved from how to define ontologies to how to reasoning with ontologies, from how to publish machine readable contents to how to integrated machine readable contents and their semantics, and from how to build proof-of-concept systems to how to develop commercial quality systems.

I’m looking forward to another five years of Semantic Web development.

Jeff Thurston at Vector One writes,

“I think interoperability is not THE issue. The issue in the GIS and geospatial arena is structural barriers internal to enterprises. It is all about breaking the barriers between people.

We need to get back to the basics, discussing what it is we want to do with the many high quality tools we have created across this industry and how they can really be used - and - changing the structural processes in organizations to make it happen.”

I think interoperability is a problem, at least from the knowledge integration point of view. Organizations have invested a lot of time and money to collect information. Different information collected over the years often is not stored in the same representation format.

Though his suggestion makes great sense, I doubt that it’s the most cost-effective solution. It’s not always economical to reconsolidate datastores just because we want different schemas to align and use shared vocabularies. NGA people told me once, “reconsolidation is expensive”.

I believe semantic interoperability is a better solution.