Building the Semantic Web for business applications

On the borders between computer science and management studies, researchers at the Erasmus Centre of Business Intelligence (ECBI) are contributing to the Semantic Web and building applications to use it for business.

According to the vision of its founder, Tim Berners-Lee, in the Semantic Web, all machines are interconnected, truly understand and reason with data, and convert it efficiently into actionable information and knowledge.

Around this topic, ECBI organised the first <link events detail _blank>International Symposium on Semantic Web Information Systems (IS-SWIS 2013). National and international experts gathered to discuss recent advances in the field and possible applications.

“Ontology” was the watchword: as models describing the world, ontologies are an important cornerstone of the Semantic Web. More specifically, they are taxonomies in which knowledge is structured and stored, using concepts and relations.

Capturing time and space in ontologies

Researchers have long been breaking their heads over how to capture time and space in ontologies – for instance indicating when and where specific events happened, or expressing different states through time. Under the auspices of a large European project, Professor Euripides Petrakis and his team at the Technical University of Crete have conducted pioneering research in this field, managing to create the required spatio-temporal support.

While very valuable to users, spatio-temporal ontologies are inevitably complex. To shield users from that complexity, researchers have added an “abstraction layer.” This layer improves user experience; in this way the team aims to pave the way to adoption by the general public.

Automatic harvesting of ontological structures

As societies change, so should the knowledge stored in ontologies. Up until now, domain experts updated ontologies manually – a tedious, repetitive, error-prone, and time-consuming job. In response, researchers like Dr Kelly Zervanou, working at Radboud University Nijmegen, are developing new mechanisms to automate this process. Their innovative research enables the semi-supervised harvesting of ontological structures.

Mapping out ontologies

Semantic Web applications each have their own knowledge bases, making it difficult for them to work together, even when they operate in the same domain. Focusing on e-commerce ontologies for products, <link people damir-vandic>Damir Vandic, PhD candidate at ECBI, is determining similarities between concepts, thereby mapping out ontologies. This mapping boosts interoperability between applications and could eventually facilitate the process of merging similar ontologies.

Extracting facts from texts

<link people frederik-hogenboom>Frederik Hogenboom, another PhD candidate at ECBI, has taken up the challenge to populate existing ontologies with facts extracted from texts. In particular, he investigates ways to extract financial events from news texts and to import them into ontologies. This data can be used in follow-up steps to process the news, but can also be useful for applications such as <link research centres business-intelligence news featuring detail>algorithmic trading. Moreover, the spatio-temporal technologies developed by Professor Petrakis are a welcome addition for future studies in this line of research.

Searching through ontologies

In the case of the Semantic Web, the human desire to record as many facts as possible results in ever richer ontologies. But searching through such ontologies – querying – is an expensive, time-consuming operation, even more so when spatio-temporal dimensions are added. Dr Jan Hidders of Delft University of Technology is developing a smart approach to indexing ontologies, in order to reduce query times. His use of structural indexes proves suitable and profitable for real-world applications.

Using semantic resources for interpretation

Once we have ontologies rich in knowledge, and we’re able to retrieve that knowledge from unstructured data, how can we use it? In his research, <link people alexander-hogenboom>Alexander Hogenboom demonstrates how semantic resources can help improve text interpretation. His focus is on determining the sentiments expressed in texts. Future research would enable the automatic understanding of people’s feelings about specific products, companies, and brands.

Conclusion

“We have come one step closer to a richer, more intelligent Web,” says Frederik Hogenboom, looking back on the symposium. The presented results are also relevant for business. Hogenboom: “Possible applications are information monitoring tools for complex systems like financial markets or markets for products and services, but also decision-support tools to be used in applications as for example algorithmic trading.”

IS-SWIS was funded by the Erasmus Research Institute of Management (ERIM) and the School for Information and Knowledge Systems (SIKS).