Controlled Natural Language meets the SemanticWeb

Controlled Natural Language meets the Semantic Web Rolf Schwitter and Marc Tilbrook Centre for Language Technology Macquarie University Sydney, NSW 2109, Australia {schwitt|marct}@ics.mq.edu.au Abstract In this paper we present PENG-D, a proposal for a controlled natural language that can be used for expressing knowledge about resources in the Semantic Web and for specifying on- tologies in a human-readable way. After a brief overview of the main Semantic Web en- abling technologies (and their deficiencies), we will show how statements and rules written in PENG-D are related to (a subset of) RDFS and OWL and how this knowledge can be translated into an expressive fragment of first-order logic. The resulting information can then be further processed by third-party reasoning services and queried in PENG-D. 1 Introduction “The Semantic Web is an extension of the current web in which informa- tion is given well-defined meaning, bet- ter enabling computers and people to work in cooperation.” (Berners-Lee et al., 2001) This claim falls short, since the languages – based on RDF – that have been designed for making statements about Web resources and for specifying ontologies are not built for hu- man consumption. These languages have been developed with machine processability and au- tomatic information exchange in mind (Manola and Miller, 2004; Smith et al., 2004). They are painful to read, write and understand for non- specialists. To overcome the disadvantages of machine- processable formal languages based on RDF and full natural languages that are far too expressive for the task at hand, we will introduce PENG- D, a controlled natural language that reconciles rigid formality and natural familiarity. In a nutshell, a controlled natural language is a subset of a natural language that has been restricted with respect to its grammar and its lexicon. Grammatical restrictions result in less complex and less ambiguous sentences, while lexical restrictions reduce the size of the lexi- con and the meaning of the lexical entries for a particular domain (Huijsen, 1998). Using a controlled language for knowledge representa- tion, specification texts become easier to read and understand for humans, and easier to pro- cess for machines (Schwitter, 2004). Traditionally, controlled natural languages have been classified into two major cate- gories: human-oriented and machine-oriented controlled natural languages. Human-oriented controlled natural languages have been designed to improve the readability and understand- ability of technical documents, particularly for non-native speakers. An example are aircraft maintenance documents in the aerospace indus- try (AECMA, 2004). Machine-oriented con- trolled natural languages have been developed to ameliorate the quality of machine translation for technical documents (Mitamura, 1999) and for writing specification texts that can be easily translated into a formal language (Fuchs et al., 1999; Schwitter, 2002; Sowa, 2004). Most of these machine-oriented controlled natural languages are defined by their trans- lation into first-order logic that automatically restricts their expressivity to a small subset of constructions compared to full English. Since it is very likely that the emerging Se- mantic Web will finally rely on a variant of description logic as knowledge representation …

Content truncated to 500 words