CCS ONTOLOGY: Domain Knowledge for Semantic Search

Short Introduction

The number of publications to CO2 capture and sequestration (CCS) increases exponentially. In order to keep track of the information overload and provide efficient support for literature search, a controlled vocabulary (CCS Ontology) has been designed that is used for automatically sorting textual content.

A utilization of the ontology within a semantic search system is being shown. Compared to conventional search engines, users of this system benefit from a considerably improved overview of search results and from the use of search-term synonyms. Thus, time can be saved and can be used otherwise.

Method and Results

The general idea of our work is an automated interaction of text mining and classification in categories using domain knowledge (ontologies) as shown by Doms (2009) for the life-sciences. On the basis of a new built CCS Ontology, text documents are being 'mined' and annotated for ~1,100 CCS-related terms and synonyms with a focus on geosciences, i.e. physical property names, parameters, monitoring methods and more. Being implemented within a semantic search engine, the CCS Ontology gives a user the opportunity to navigate through hierarchically categorized search results. Text content with synonym-terms will be found without having explicitly searched for the synonym and hence makes the literature research more effective and efficient.

Outlook

Word sense disambiguation remains to be a non-trivial task. Furthermore, new evaluation methods need to be developed (Renear, 2009) since traditional approaches like retrieval precision and recall lack in providing the kind of analysis needed to fully show the overall advances of the efficient search as shown here. However, we are confident that the CCS Ontology could also be used not only to sort textual content but also to link to matching database content of (e.g.) subsurface models as considered in the CO2-MoPa project (Thomsen, 2010).

Carbon Capture and Sequestration (CCS) related publication count per year shows exponential increase (green solid line) while selected geophysical search terms are related to constant or only slightly increasing publication numbers (black markers). The aim is to support scientific readers with a semantic search system that automatically ‘understands’ textual content and classifies/sorts search results on based on a new CCS Ontology in order to save time spent on manual search, understanding and classification.

The CCS Ontology consists of concepts and relations between them (left). A concept in this ontology can have synonyms and also a definition. Synonyms are used in order to find similar or equal results to the user’s query without having explicitly searched for the synonym terms. Thus, results with a semantic search engine (right) are more complete than with a ‘classic' one.

References:

DOMS, A. (2009): GoPubMed: Ontology-based literature search for the life sciences, Ph.D. thesis, Department of Computer Science, Technical University of Dresden.

GUTKNECHT, B. D. (2010): Zur Ontologischen Suche bei Recherchen in der CCS Literatur, diploma thesis, Institute of Geosciences, Christian-Albrechts-University Kiel.

RENEAR, A.H., Palmer, C.L. (2009): Strategic Reading, Ontologies, and the Future of Scientific Publishing, Science, 325, 828-832.

THOMSEN, A. et al. (2010): On the way to synoptic interpretation of geoscientific data in joint CCS project CO2-MoPa, Talk, Geoinformatik 2010, Kiel.

Partners:

Transinsight GmbH, Dresden

http://www.transinsight.com

Potential Geo Services

http://www.potentialgs.com

Contact:

B. D. Gutknecht, Email: benjamin[at]geophysik.uni-kiel.de

H.-J. Götze, Email: hajo[at]geophysik.uni-kiel.de

M. R. Alvers, Email: malvers[at]transinsight.com