Information Extraction from Unstructured Texts by Means of Syntactic Dependencies and Constituent Trees
Raoul Schönhof, Axel Tenschert, Alexey Cheptsov · 2015
This paper presents a technology of automated knowledge extraction from unstructured text corpora by leveraging computer linguistic tools and cross-fertilizing them with the semantic ontologies techniques. In our approach, the quality of information (e.g., in form of OWL ontologies) that is derived by semantic analysis techniques from large domain- specific text corpora can be considerably improved by incorporating linguistic analysis tools that help gain a deeper insight into the grammatical structure of the analysed texts and thus allow the reasoning engines to cover a much wider set of rules and patterns, also positively impacting the performance. The novelty of our approach lies in a possibility of its application to the domains that require a very high quality of the knowledge extraction and analysis, such as reasoning for legacy data collections. We propose a system architecture for the implementation of our approach and illustrate its use on a practical use case for legislative and regulatory information analysis. Keywords-Knowledge Representation; Legal Systems; Ontology; OWL; Big Data; Reasoning; DreamCloud Project.