Natural Language Processing Techniques for Document Classification in IT Benchmarking Automated Identification of Domain Specific Terms

摘要

In the domain of IT benchmarking collected data are often stored in natural language text and therefore intrinsically unstructured. To ease data analysis and data evaluations across different types of IT benchmarking approaches a semantic representation of this information is crucial. Thus, the identification of conceptual (se-mantical) similarities is the first step in the development of an integrative data management in this domain. As an ontology is a specification of such a conceptualization an association of terms, relations between terms and related instances must be developed. Building on previous research we present an approach for an automated term extraction by the use of natural language processing (NLP) techniques. Terms are automatically extracted out of existing IT benchmarking documents leading to a domain specific dictionary. These extracted terms are representative for each document and describe the purpose and content of each file and server as a basis for the ontology development process in the domain of IT benchmarking.