论文部分内容阅读
It is widely recognised that there is a tension in functionally-oriented register analysis between,on the one hand,the desirability of identifying relevant patterns of linguistic choices in as large a corpus as possible and,on the other,the need to include choices which the computer software currently available is not yet sophisticated enough to recognise automatically.The project that I am engaged in points up this tension particularly sharply.My aim is to explore differences and similarities in the deployment of the resources of conjunction in different registers in English.In this article,I begin by outlining the model of conjunction with which I am working.I then address an issue that I see as increasingly urgent if corpus linguistics is to continue setting the research agenda as it has done in the past decades:that is,how far automatic recognition of fundamental features of discourse can be pushed,particularly in the direction of functional realizations.Having summarised the main kinds of problems that appear to make automatic recognition difficult,or perhaps even impossible,in this area,I focus on an example that at first sight is relatively straightforward:relations signalled by the conjunction because.I discuss analytical difficulties relating to the classification of the different kinds of relationships found.In particular,I explore the possibility of establishing reliable,automatisable tests for discriminating between instances of conjunctive relationships which are close in some way(such as where the same closedclass conjunctive signal occurs in different relationships).A distinction is drawn between signals and indicators of relationships,and I put forward suggestions for the kinds of capabilities that need to be developed in computer programs in order to handle functional realizations more effectively and accurately than is possible at present.
It is widely recognized that there is a tension in functionally-oriented register analysis between, on the one hand, the desirability of identifying relevant patterns of linguistic choices in as large a corpus as possible and, on the other, the need to include choices which the computer software currently available is not yet sophisticated enough to recognize automatically. The project that I am engaged in points up this tension was sharply. My aim is to explore differences and similarities in the deployment of the resources of conjunction in different registers in English. In this article, I begin by outlining the model of conjunction with which I am working. I then address an issue that I see as per urgent if corpus linguistics is to continue setting setting the research agenda as it has done in the past decades: that is , how far automatic recognition of fundamental features of discourse can be pushed in, particularly in the direction of functional realizations. kinds of problems that appear to make automatic recognition difficult, or perhaps even impossible, in this area, I even on an example that at first first sight is relatively straightforward: relations signalled by the conjunction because.I discuss analytical difficulties relating to the classification of the different kinds of relationships found.In particular, I explore the possibility of establishing reliable, I automatisable tests for discriminating between instances of conjunctive relationships which are close in some way (such as where the same closed class conjunctive signal occurs in different relationships). A distinction is drawn between signals and indicators of relationships, and I put forward suggestions for the kinds of capabilities that need to be developed in computer programs in order to handle functional realizations more effectively and accurately than is possible at present.