Natural Language Generation: New Results in Artificial Intelligence, Psychology and Linguistics
Gerard Kempen · Medical Entomology and Zoology · 2011
Language generation is beginning to attract the attention of a large number of researchers. Generators are developed in order to try out the theories of their authors, or to provide an output facility for increasingly sophisticated language processing systems. The pioneering generators of the 70's and early 80's often employed simple models of language and syntax, and concentrated on specific tasks ? Winograd's (1972) SHRDLU, for example, used templates for output generation, Davey's (1979) PROTEUS concentrated on specific words and sen tence structures, and Simmons and Slocum (1972) mapped portions of semantic networks into sentences. Researchers are now taking a broader view, and are starting to draw conclusions from those experiments. Better developed linguis tic models and methods are being used to produce more accurate and flexible texts, and new aspects of generation are being tackled. More interest is being taken in efficiency, portability, coverage and psychology. These concerns are all apparent in this book, which gives an overview of the current state of research on generation. It is a compilation of edited versions of all 29 papers presented at the Third International Workshop on Natural Language Generation, which took place in the Netherlands in August 1987. There are contributions from almost all the leading researchers in the field, covering a wide range of topics. The relevant disciplines of linguistics, psychology and computer science are all represented. Since attention focuses on text, works related to speech and speech synthesis are absent. To inject some order, Kempen has classed the entries under six headings, based on content, rather than discipline ? Pragmatic Aspects, Generation of Connected Discourse, Generator Design, Grammars and Grammatical Formalisms, Stages of Human Sentence Production and Aspects of Lexicalisation. Some of the contributions describe established generation models. Mann and Thompson, for example, describe their discourse planning technique, Rhetorical Structure Theory, and Hovy presents the discourse planner PAULINE, which is sensitive to pragmatic goals. Paris and McKeown report on an extension to another well known model (McKeown, 1985), on which the TEXT system