International Journal of Corpus Linguistics - Volume 4, Issue 1, 1999

This paper reports a study of the phenomenon of successful turn-bidding (STB) in a corpus of English conversation taken from the London-Lund Corpus of spoken English. It analyses the features of the language environment associated with turn-bidding speech acts and the linguistic devices which are used by the turn-bidders. It also examines the effect of different speech domains, different speakers' relationships, social status, and gender on the incidence of successful turn-bidding. Finally, consideration is given to the importance of successful turn-bidding strategies for the learning and teaching of English by speakers of other languages.

The British National Corpus (BNC) contains a spoken component of about 10 million words, consisting of spoken language of various kinds produced by different speakers in a variety of situations. Starting from an end-user s perspective, this paper surveys the potential of this resource and some possible problems one might encounter if not fully versed in the details of the compilation and coding plans. Among the issues touched upon are questions relating to the composition of the component, the transcription principles employed, and points relating to the nature and coverage of the mark-up. By way of illustration, examples are drawn from a case study of the variant forms gonna and going to.

Corpus linguistics is being used for a wide range of research tasks. A database on English lexicology was compiled on the basis of the semantic, syntactic, and pragmatic information found in the stylistically marked lexical items of the COBUILD 1987. It offers linguistic information on 7981 units annotated as "formal" or "informal" in the COBUILD 1987. This database may be used for quantitative and qualitative analysis and for a critical evaluation of the classifications used in the COBUILD dictionary. The aim of this paper is basically descriptive: (i) some information is given on the linguistic information stored; (ii) some tentative conclusions are drawn. For example, if the COBUILD formality labels were assigned rigorously, it might be concluded that (i) "formality" is a skew system; (ii) that the "formal/informal" scale has an equi distribution; (iii) that emphasising informal adverbs are so numerous because they illustrate a tendency to exaggeration observed in informal, relaxed situations, in which interactants usually struggle for controlling the message.

In this article I argue that, from a methodological point of view, descriptive studies improve considerably if they use a multi-method approach to the data, more specifically, if they use a combination of corpus data and experimental data. In the modern conception of corpus linguistics, intuitive data play an important role. The linguist formulates research hypotheses based on his or her intuitive knowledge. These hypotheses are then tested on the corpus data. I argue that a sound descriptive study should not end with simply stating the results from the corpus study. Instead, the corpus data have to be supplemented. An appropriate way to supplement corpus data is through the use of elicitation techniques. I illustrate the multi-method approach on a case study of floating postmodification in the English noun phrase.

The paper discusses the design of a new computational model based on corpora—the Structural Boundary Model (SBM), particularly for the purpose of NLP. The Structural Boundary Model is constructed on the basis of parsed corpora. It consists of two main bodies, namely structural boundary data and CFG rules. The grammar supports parsing in a unique way by assigning structural boundary labels retrieved from a parsed corpus as a training corpus for the parser. Parsing experiments have demonstrated that the Structural Boundary Model is an appropriate novel computational model for parsing.

The following article is meant to discuss the status of corpus linguistics, how it is seen and sees itself as a field: Is it merely a method of doing linguistics, or can it be considered a distinct approach to language description? In our argument, we claim that corpus linguistics is on the way of becoming more than a methodology, since its research results are increasingly interpreted with regard to their impact on the commonly held views about language. Dealing with these interpretations, we have noticed a number of similarities with assumptions made by cognitive linguistics, and we aim at showing that the two trends—corpus linguistics and cognitivism—are compatible in that they complement each other.

The review describes the design and features associated with two Windows text analysis programs: a concordancer MonoConc and a parallel concordance program ParaConc. The general operation and potential uses of the two programs are briefly explored.