© Bilfinger SE

Passing on expert knowledge and experience stemming from projects is, and has always been, one of Bilfinger’s key corporate tasks.  A semantic search system was developed in collaboration with scientists from Fraunhofer-Gesellschaft that links various applications with just one query, sifts through all available information resources, is operable intuitively and then issues structured results. This R&D project undertaken by Bilfinger provided crucial findings in developing the market maturity of CONWEAVER’s search technology.

Such infrastructural projects as a six-lane elevated road in Bangkok, ensuring water supply in Australia for instance, sustainable structural engineering, intelligent service concepts for conventional and nuclear power plants as well as projects in accomplished by the approximately 60,000 employees of Bilfinger SE worldwide within the last months and years. Around the globe this well-known company offers tailored solutions in real estate infrastructure, energy technology and industrial services to its customers. Hence, complex projects requiring vast knowledge and experience are everyday business. “Our employees gain know-how with every project. Making this expert knowledge accessible for everybody is elementary for our company, but yet not easy to accomplish”, says Beate Kögel, Technical knowledge management project manager at Bilfinger SE.

Our employees gain know-how with every project. Making this expert knowledge available for everybody is elementary for our company, but yet not easy to accomplish.

Beate Kögel, Technical knowledge management project manager at Bilfinger SE

While the employees had access to project information, contacts and manuals through a technical portal as well as norms and monographs through a library system, a link between the different data sources did not exist. Comprehensive queries were impossible; again and again posing the same question to project engineers and employees in which of the numerous data sources to find the desired information. “In addition, each data base needed to be operated differently, making it necessary for our employees to become acquainted with new query languages over and over again”, explains Kögel. Often enough, such research was unsuccessful and aborted with frustration – too complex, too lengthy, too little benefit.

So much for the initial situation in 2004, when Beate Kögel contacted Dr. Thomas Kamps, then Head of Intelligent Content Technologies at Fraunhofer IPSI. “Our goal was to develop a semantic search function that integrated the data of our 16 different data bases, with all data sources intelligently linked. Dr. Kamps of Fraunhofer-Gesellschaft has been the ideal partner for this complex design, since, at the time, he had more than ten years of practical experience in the field of knowledge technology”, explains the project manager.

The Task:Connect data, simplify the search

In a joint project group comprising users and scientists, the requirements were discussed and defined. Key criteria were soon obvious:

From discussions with staff, it became apparent that the system would only be accepted if it was absolutely easy and intuitive to use. No matter how good the technology, as long as it is not as easy to use as Google, it will not stand a chance in day to day business. Tediously familiarizing oneself into new software was out of question.

Dr. Thomas Kamps, CEO of CONWEAVER GmbH

That said, the project posed numerous challenges to the development team led by Dr. Kamps. Even though all data sources were related in terms of content, they were not linked technically. The data bases consisted of Lotus Notes and SQL applications with differing structures and languages. To manage such diversity was not easy. “At first, we connected existing structured data, which we then transformed into structured data with the help of statistical and linguistic methodologies, then feeding them into the network. In addition, the semantic network should obviously be kept up to date. Thus, for example, the index must be updated when new documents are added to the original system. Everything automatically, of course”, as Kamps explains the work that had to be done by him and his team.

What sounded relatively easy did pose an enormous task for the system, which was solved by approximately 200 analysis modules. These analysis modules were assembled into specific, company-related workflows that analyze, structure, translate and link the data. At the end of this process is a knowledge network that is automatically generated and updated, spanning the entire integrated database and facilitates a semantic, i.e. content-based, search.

The Result: Globally linked, bilingual, clearly arranged

Two years later, the development process was completed and the system went into use. “We created a knowledge network that automatically generates itself. Through the key database developed by the then Bilfinger Berger AG in German and English, which considered semantic concepts and is connected to the knowledge network, the system may be used worldwide and multilingually. Furthermore, we integrated mechanisms that consider variations within formulation of queries”, explains Kamps. This means that German inquiries also display English hits and that, for instance, a query for methods for water treatment also delivers hits for water processing plants.

That different designations for the same issue are considered within the same inquiry versus a separate search for each term, facilitates the search for our staff tremendously and saves a lot of time

says Beate Kögel. Unlike traditional internet search engines, the results are aggregated across data sources. “As an engineer at Bilfinger SE looks for Brandschutz im Tunnelbau, tore searches for all corresponding experts, among all other things. The system not only ‘knows’ that fire protection and tunneling are the translations for Brandschutz and Tunnelbau. It is also able to detect the term in reports, which, in turn, are assigned to projects. Since the project, report and project manager are stored within the project’s database, its manager may be proposed as an expert through the knowledge network’s linkage”, stresses Kamps as the crucial advantage for tore as a semantic search technology. 

To meet the demand for a clearly arranged hit list, the results are displayed by expert, project, specialized information/manual, library entries as well as Research and Development (R&D). What is more, it is possible to go into further detail. For example, all documents an expert has issued are displayed and may be opened.

The tore search is now used worldwide at Bilfinger. To demonstrate the technology’s commercial success, various value-adding indicators were identified across selected departments, with unambiguous results: “With tore it is now possible to obtain information as well as to improve our staff’s decision-making and knowledge significantly. Even complex questions can now be answered in a short time and experts traced quickly. Spatial boundaries are no longer relevant. Overall, our competitive position has improved through the use of the semantic knowledge network”, Beate Kögel summarises. tore’s first implementation (together with K-Infinity by intelligent views) delivered valuable information for the conception of optimized data analysis and search technologies. This information constitutes the foundation for the development of CONWEAVER software. tore has been running at Bilfinger SE since December 2008 on the basis of CONWEAVER. Along with the conversion to CONWEAVER, additional specific data resources from structural engineering were integrated in the semantic search.

Bilfinger SE’s tore  search engine, using CONWEAVE software, takes a completely different approach compared to other current search engines. Based on the specific corporate data signature, an optimum of knowledge is extracted, unstructured data converted to structured data while organizing the whole in a cohesive knowledge network.


Your Contact

Sebastian Dörr