The research for this thesis has been concerned with defining and demonstrating the existence of certain semantic elements in English natural language science text which can be called metainformation. Meta-information is described as being the organisational-, rather than the conceptual properties of an author's 'message' in text. Conceptual information is that subject-related output from a document which readers assimilate or synthesise with their current state-of-knowledge. Meta-information reflects the organisation or structural format used by an author to present conceptual information for transfer from text to readers. The example used here to demonstrate the existence of meta-information, is a format for the presentation of empirical argument in science text. At its most simplep a meta-informational element could be a report section-heading like, INTRODUCTION, which describes (we assume), the contents of the subsequent text. At a lower level of analysis the phrase, 'This paper describes contains some semantic inference that the complete statement is one of an introductory nature; thereforep such a statement could be labelled as one of INTRODUCTION for meta-informational purposes. A 'grammar' or set of meta-informational elements, has been developed as a means of identifying certain semantic aspects of text. This grammar is based on some experimental evidence and the consensus view of readers and writers of science text who produced what has been called a conventional format for empirical argument presentation. An initial set of rules for implementing this grammar have also been developed. The rules have been tested for replicability with positive results. Although analysis of full text hasshown deviation from a 'conventional argument structure readers' summaries of the same text conform to this structure. Thus, a model of the phenomenQn of information transfer from text to readers, which includes a structural transformation process based on the experimental results, has been built. A computer simulation is given to demonstrate the model in an inter-active program-user system designed to produce summaries of whole text. The thesis is that evidence exists for the presence of meta-information in science text and that if a grammar appropriate to the kind of output information required by users is built, highly structured text could be produced so that the process of information transfer is optimised.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:471466 |
| Date | January 1979 |
| Creators | Sallis, P. J. |
| Publisher | City University London |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://openaccess.city.ac.uk/8580/ |
Page generated in 0.0019 seconds