Sourcing and procurement are mayor tasks of supply chain management, which increasingly benefit from Web technologies in that producers respectively service suppliers make information on their services retrievable online for the prospective consumers. Thus, especially for small and medium enterprises, the Web acts as an intermediary to attract new consumers from all over the world. The wide range of branches, the degree of specialization, as well as the size and resources of service supplying companies provoke, that their Web sites promoting their services highly vary in extent, detail and trendiness. In spite of the improved accessibility of information these heterogeneities and the mere number of existing Web sites entail that a search for manufacturers, offering exactly the service the consumer is in demand for, turns out to be a very tricky and time consuming task. Information extraction technologies constitute the prerequisite to enable the semi-automatic support for finding the best fitting service supplier.

MARLIES is an information extraction system targeted at extracting contact data and detailed data on services offered by companies in the manufacturing industry via their Web sites. The system constitutes one component of the intermediary system tech2select, which is operated by the Austrian company Tech2select GmbH. Its development by FAW is partly funded by the Austrian Research Promotion Agency FFG under grant FFG 817789.

Services as treated by MARLIES are machines or manufacturing processes promoted by a supplier. In order to build the base for a business assignment the services must be specified in sufficient detail, including information on the processable material and dimension, which are further described by measurements, units and values. On account of the given requirements an ontology and rule based approach was implemented. The major challenges of MARLIES are on the one hand the proper ontological modelling of highly structured and complexly related technical data, which constitutes the basis for the realized approach of an ontology aware annotation, on the other hand, the extraction of relations between the data units while tackling structural provocations, as related data on a service might be spread over several Web pages and is often concealed in nested tables.