Dorte Haltrup Hansen

Ontology-based question answering in a federation of university sites: The MOSES case study

This paper deals with a new approach to ontology-based QA in which users ask questions in natural language to knowledge bases of facts extracted from a federation of Web sites and organised in topic map repositories. Our approach is being investigated in the context of the EU project MOSES.

Language technology in knowledge-organization systems

This paper describes the language technology methods developed in the Danish research project VID to extract from Danish text material relevant information for the population of knowledge organization systems (KOS) within specific corporate domains. The results achieved by applying these methods to a prototype search engine tuned to the patent and trademark domain indicate that the use of human language technology can support the construction of a linguistically based KOS and that linguistic information in search improves recall substantially without harming precision (near 90%). Finally, we describe two research experiments where (1) linguistic analysis of Danish compounds and is exploited to improve search atrategies on these (2) linguistic knowledge is used to model corporate knowledge into a language-based ontology.

Anonymization of Court Orders

We describe an anonymization tool that was commissioned by and specified together with Schultz, a publishing company specialized in Danish law related publications. Unavailability of training data and the need to guarantee compliance with pre-existing anonymization guidelines forced us to implement a tool using manually crafted rules. We used Bracmat, a programming language that is specialized in transforming tree data structures, to meet the requirement to pass the XML structure of the input document unscathed through the whole workflow. The tool attains a reassuringly good recall, makes almost no chunk errors and reduces the found entity designators to a nearly correct set of entities that the input text refers to, minimizing the time needed for manual check and post-editing.