An Update on PENG Light

An Update on PENG Light Colin White Department of Computing Macquarie University Sydney NSW 2109, Australia colcwhite@gmail.com Rolf Schwitter Centre for Language Technology Macquarie University Sydney NSW 2109, Australia Rolf.Schwitter@mq.edu.au Abstract This paper presents an update on PENG Light, a lightweight and portable controlled natural language processor that can be used to translate a well-defined subset of En- glish unambiguously into a formal target language. We illustrate by example of a Firefox extension that provides a simple in- terface to the controlled natural language processor how web pages can be annotated with textual information written in con- trolled natural language and how these an- notations can be translated incrementally into first-order logic. We focus in partic- ular on technical aspects of the controlled language processor and show in detail how look-ahead information that can be used to guide the writing process of the author is generated during the parsing process. Ad- ditionally, we discuss what kind of user interaction is required for processing un- known content words. 1 Introduction Computer-processable controlled natural lan- guages are well-defined and tractable subsets of natural languages that have been carefully de- signed to avoid constructions that may cause am- biguities (Fuchs et al., 1998; Schwitter, 2002; Sowa, 2004; Barker et al., 2007). Instead of encoding a piece of knowledge in a formal lan- guage that is difficult to understand for humans, a controlled natural language can be used to ex- press the same information in a direct way using the vocabulary of the application domain. There is no need to formally encode this information since a computer-processable controlled natural language can be translated automatically and un- ambiguously into a formal target language by a machine. This has the advantage that everybody who knows English can understand a text writ- ten in controlled natural language and that a ma- chine can process this text since it corresponds to a formal notation. In order to support the writ- ing of these texts, text- and menu-based predic- tive editing techniques have been suggested that guide the writing process of the author (Tennant et al., 1983; Schwitter et al., 2003; Thompson et al., 2005; Kuhn, 2008). These techniques give the author a way to match what he or she wants to express with the processing capabili- ties of the machine and result in user-friendly and self-explanatory interfaces. However, prac- tically no details have been published so far how these predictive techniques can actually be imple- mented in a controlled natural language proces- sor that processes a text incrementally. In this pa- per, we will make up for this neglect and show how these predictive techniques have been imple- mented for the controlled natural language pro- cessor of PENG Light and discuss how the lan- guage processor uses these techniques while com- municating over an HTTP connection with a sim- ple AJAX-based tool designed for annotating web pages in controlled natural language. The rest of this paper is structured as follows: In Section 2, we give a brief overview …

Content truncated to 500 words