Karla A. V. Borges

Ontologies and Knowledge Sharing in Urban GIS

Data and knowledge exchange among users of urban information systems presents many challenges. This paper discusses issues related to the use of ontologies in the development of urban geographic information systems and proposes the creation of software components from diverse ontologies as a way to share knowledge and data. These software components are derived from ontologies using an object-oriented mapping. The translation of an ontology into an active information system component leads to ontology-driven information systems and, in the specific case of geographic applications, to ontology-driven geographic information systems. We analyze the urban environment from the ontologists point of view and make some inferences about the relationship between knowledge and data sharing, and the theory of bona fide and fiat objects. We also discuss implementation issues, such as the use of Ontolingua as an ontology editor, and CORBA IDL generator, CORBA, and Java as object platforms.

OMT-G: An Object-Oriented Data Model for Geographic Applications

Semantic and object-oriented data models, such as ER, OMT, IFO, and others, have been extensively used for modeling geographic applications. Despite their semantic expressiveness, such models present limitations to adequately model those applications, since they do not provide appropriate primitives for representing spatial data. This paper presents OMT-G, an object oriented data model for geographic applications. OMT-G provides primitives for modeling the geometry and the topology of spatial data, supporting different topological structures, multiple views of objects, and spatial relationships. OMT-G also includes tools to specify transformation processes and presentation alternatives, that allow, among many other possibilities, modeling for multiple representations and multiple presentations. In this way, it overcomes the main limitations of the existing models, thus providing more adequate tools for modeling geographic applications. A comparison with other data models is also presented in order to stress the main advantages of OMT-G.

Discovering geographic locations in web pages using urban addresses

This paper presents an approach that helps to discover geographic locations from the recognition, extraction, and geocoding of urban addresses found in Web pages. Experiments that evaluate the presence and incidence of urban addresses in Web pages are described. Experimental results, based on a collection of over 4 million documents from the Brazilian Web, show the feasibility and effectiveness of the proposed method.

Spatial data integrity constraints in object oriented geographic data modeling

An important activity in the design of a particular database application consists in identifying the integrity constraints that must hold on the database, and that are used to detect and evaluate inconsistencies. It is possible to improve data quality by imposing constraints upon data entered into the database. These constraints must be identified and recorded at the database design level. However, it is clear that modeling geographic data requires models which are more specific and capable of capturing the semantics of geographic data. Within a geographic context, topological relations and other spatial relationships are fundamentally important in the definition of spatial integrity rules. This paper discusses the relationship that exists between the nature of spatial information, spatial relationships, and spatial integrity constraints, and proposes the use of OMT-G, an extension of the OMT model for geographic applications, at an early stage in the specification of integrity constraints in spatial databases. OMT-G provides adequate primitives for representing spatial data, supports spatial relationships, and allows topological, semantic and user integrity rules to be specified in the database schema.

Semantic Expansion of Geographic Web Queries Based on Natural Language Positioning Expressions

The need for better Web search tools is getting increasing attention nowadays. About 20% of the queries currently submitted to search engines include geographic references. Thus, it is particularly important to work with the semantics of such queries, both by understanding the terminology and by recognizing geographic references in natural language text. In this paper, we explore the use of natural language expressions, which we call positioning expressions , to perform geographic searches on the Web, without resorting to geocoded data or gazetteers. Such positioning expressions denote the location of a subject of interest with respect to a landmark. Our approach leads to a query expansion technique that can be explored by virtually any keyword-based search engine. Results obtained in our experiments show an expressive improvement over the traditional keyword-based search and a potential path for tackling many kinds of common geographic queries.

The role of gazetteers in geographic knowledge discovery on the Web

The Web is a large source of geographic information. Many Web documents have one or more spatial references, such as place names, addresses, zip codes or phone numbers. These spatial references are usually found in a semistructured fashion, which allows humans to identify and assign a geographic meaning to documents. In this paper, we discuss the important role that gazetteers, which are spatial catalogues of place names, can play in automating this process, and introduce the Locus gazetteer. Locus has been designed to hold not only place names for entities such as cities and rivers, but also to handle intra-urban place names, such as street names, urban landmarks, and postal addresses, along with their spatial relationships, through an ontology of places. We demonstrate that ontologically-enhanced gazetteers, such as Locus, are very useful for discovering the geographic context present on Web pages, and are often used in many other applications, such as in address geocoding for geographic information systems. To efficiently accomplish these tasks, the gazetteer must have a large database of spatial references; however, such a database is hard to obtain in emergent countries such as Brazil, in which available official geographic databases are limited and not well updated. As a way to tackle this problem, we describe a semi-automatic method used to populate the Locus gazetteer with geographic content extracted directly from the Web. To evaluate our work, an experiment was conducted, focusing on testing the Locus gazetteer data quality and comprehensiveness.

Geographic web search based on positioning expressions

In this paper, we explore the use of natural language expressions to perform geographic searches on the Web, without resorting to geocoded data. Such expressions denote the positioning of a subject with respect to a landmark --- a valuable source of geographical context embedded in the unstructured text of Web documents. Our approach leads to novel query expansion techniques that can be explored by virtually any keyword--based search engine.

Ontology-driven discovery of geospatial evidence in web pages

When users need to find something on the Web that is related to a place, chances are place names will be submitted along with some other keywords to a search engine. However, automatic recognition of geographic characteristics embedded in Web documents, which would allow for a better connection between documents and places, remains a difficult task. We propose an ontology-driven approach to facilitate the process of recognizing, extracting, and geocoding partial or complete references to places embedded in text. Our approach combines an extraction ontology with urban gazetteers and geocoding techniques. This ontology, called OnLocus, is used to guide the discovery of geospatial evidence from the contents of Web pages. We show that addresses and positioning expressions, along with fragments such as postal codes or telephone area codes, provide satisfactory support for local search applications, since they are able to determine approximations to the physical location of services and activities named within Web pages. Our experiments show the feasibility of performing automated address extraction and geocoding to identify locations associated to Web pages. Combining location identifiers with basic addresses improved the precision of extractions and reduced the number of false positive results.

Domain and Model Driven Geographic Database Design

After many years of research in the field of conceptual modeling of geographic databases for Geographic Information Systems, experts have produced many different alternatives of conceptual data models from extensions of the Entity-Relationship model or of Unified Modeling Language (UML). However, the lack of consensus on which is the most suitable one for modeling applications in the geographical domain brings up a number of problems for field advancement, mainly problems of interoperability of database design and CASE tools. The Model Driven Architecture (MDA) approach allows the development of systems from an abstract view until the corresponding implementation code that can be automatized by means of models transformation. A UML Profile is an extension mechanism of UML which allows a structured and precise extension of its constructors, being a good solution to standardize domain-specific modeling, as it uses the entire UML infrastructure. This chapter describes the use of MDA approach in the design of databases in geographical domain; using a UML Profile called GeoProfile aligned with international standards of ISO 191xx series. The chapter also shows that with the automatic transformation of models it is possible to achieve the generation of scripts for spatial databases from a conceptual data schema in a high level of abstraction.

Searching for metadata using knowledge bases and topic maps in Spatial Data Infrastructures

Concern for environmental issues has become a reality in all sectors of society, mainly among researchers and professionals who work directly with environmental status. In this context, several studies have been undertaken on sustainable development of the Brazilian Amazon, generating a large amount of data and information. Environmental area characterization involves the knowledge about their natural, economic and social resources, as well as understanding the interaction and correlation among them. Such interdisciplinary character requires new solutions for knowledge representation. This study proposes to minimize metadata recovery problems in Spatial Data Infrastructures by using Topic Maps and Thesaurus. This approach applied to an interface aims to allow users to visually recover information from metadata catalogs.