=Paper=
{{Paper
|id=Vol-1265/pd3
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-1265/owled2014_submission_7.pdf
|volume=Vol-1265
}}
==None==
https://ceur-ws.org/Vol-1265/owled2014_submission_7.pdf
Ontology Matching for Geospatial Domain
Ujwala Bharambe ,Surya Durbha
Indian Institute of Technology Bombay. India
ujwala.bharambe@iitb.ac.in, sdurbha@iitb.ac.in
Abstract. Geospatial information is collected from different sources with disparate
terminologies resulting in highly heterogeneous information thus making its com-
pilation and retrieval difficult. Harmonization of these heterogeneities is needed for
interoperability and seamless access to the information sources by providing a
common platform for facilitating information exchange. This can be achieved by
ontology matching. This research proposes a conceptual framework for Geo-
ontology matching to tackle the two major issues for geo-information fusion: hete-
rogeneity and uncertainty by using information theoretic approaches for
representing features of ontology. Further, the multi-objective optimization tech-
nique such as Pareto Ranking technique is adapted for the optimization of the best
match pairs to derive the correspondence between two Geo-ontologies.
Keywords: Interoperability, Geo-ontology Matching, Information Theory, Pareto
ranking
1 Introduction
Interoperability in geospatial domain has garnered lot of interest in recent years with a
lot of challenges faced towards the integration of diverse geospatial information sources.
Several syntactic and structural approaches for integration are currently available, such
as the Open Geospatial Consortium’s (OGC) standards for geospatial web services.
However, these standards only facilitate interoperability at the syntactic level, which
impedes the seamless access of geospatial data. Building a semantic layer on top of the
syntactic representation enables effective integration of the semantics of different
geospatial data thereby removing gaps in the information acquisition and understanding
of the integrated data, it’s sharing and use [2]. The above can be effectively achieved by
using ontologies [1] where each source is represented by an ontology that explicitly
represents the implicit concepts of that domain. The aim of this research is to develop a
framework for geospatial ontology matching to achieve semantic interoperability thus
providing an integrated view of the geo-information sources.
2 Ontology Matching in the Geospatial Domain
The main objective of ontology matching is to resolve the discrepancy among the on-
tologies which are described by different communities using different vocabularies and
different perspectives/contexts. The difficulties in Geo-ontology matching may also be
aggravated due to peculiarities of geospatial information (i.e. topology, dimension and
orientation, shape, size and location) and due to lack of widely accepted Geo-ontology
model causing the Geo-ontologies to be at different granularity levels.
The proposed system is composed of a four components as shown in figure 1. The
first component (Geo-ontology access and analysis) takes two input ontologies and per-
forms parsing of the ontologies using OWL API. The second component (i.e. Geo-
ontology Feature Extraction) is responsible for capturing all lexical, structural and in-
stance based features of the given ontologies.
Two Input Geo-Ontologies Os and Ot Geo-Ontology Access and Preprocessing
Tree based representation of Os and Ot
Geo-ontology Feature Extraction
Lexical, Structural and Instance based Features
Geo-Ontology Analysis and Matching
Probable Match Pairs
Mapping, Selection and Interpretation using Multi-objective optimization technique
Final Selected Match Pairs
Fig. 1. Proposed System Architecture for Geo-ontology Matching
The third component (Geo-ontology Matching) comprises of the matching modules. The
matching process is divided into three main modules: 1) Lexical Matcher: This is ele-
ment based matcher (local). It uses different similarity metrics (string similarities: Jaro
Wrinker and Wordnet Similarity) to handle the high terminological (lexical) heterogenei-
ty of ontologies. 2) Structural matcher takes advantage of quantized features, which
represent the structural characteristics of ontology. Ontology matching can be seen as an
operation that takes two graph-like structures and produces a mapping between elements
of the two graphs that correspond semantically to each other. Gaussian similarity meas-
ures (eq 1) are used for calculation of the structural similarity at the local
𝑑2
el. 𝑆𝑖𝑚gauss 𝑒1, 𝑒2 = 𝑒 −2𝜎 2 (1)
Where d is Euclidean distance between relative entropy concepts [3] and 𝜎 is scaling
factor. 3) Instance base Matcher: Cosine similarity measures are used for instance based
matching. The final component is Best Match Pairs Selection and Interpretation where
bi-objective optimization problem is solved using Pareto optimization technique wherein
both the lexical and structural features are taken into equal consideration thereby achiev-
ing a most reasonable alignment and maximizing the number of correspondences in
matching. The proposed framework of Geo-ontology matching is being implemented by
using Java programming environment. The proposed framework for Geo-ontology
matching would be evaluated in the context of rapid decision making for disaster man-
agement system especially flood management. The major focus is on developing a gene-
ralized adaptive multi-strategy Geo-ontology matching system.
References:
1. Kuhn, W.: Geospatial Semantics: Why, of What, and How? Journal on Data Semantics III.
Springer Berlin Heidelberg. 3534: 1-24 (2005)
2. Lutz, M., E. Klien.: Ontology‐based retrieval of Geographic Information. International Jour-
nal of Geographical Information Science 20(3): 233-260. (2006)