Mapping key words to problem and solution

Editors’ introduction Scott’s paper shares with those of Coulthard and Hunston in this volume a focus on patterns which in different ways depend on repetition across rather than within texts. However, unlike them his concern is not with patterns of wordings but with patterns of content. He explores aspects of the ‘aboutness’ of texts, working with the notion of keyness.

Keyness relates to the frequency of particular lexical items within a text as compared with their frequency in a reference corpus. Like Hunston, Scott uses the corpus as a way of simulating the accumulating sum of exposure to our language in use that each of us experiences, as listener, reader, speaker and writer. The corpus allows him to measure the frequency with which we are likely to come across a particular word in general, and to compare this with the frequency of that word in a specific text. If the word occurs significantly more often in a text than one would predict on the basis of its frequency in the language as a whole (represented by the corpus), it should reflect some distinctive aspect of what that text is about. The word is then one of the key words of that text; and other texts in which the same word is key are likely to share concerns with the text (especially if the texts have more than one key word in common).

Scott uses this insight to explore the behaviour of a set of words which have a particular function in texts. These are words that can signal the presence of a Problem-Solution pattern: not only problem and solution but also words like unfortunately. The initial analysis of a large body of texts is done automatically; and scrutiny of the examples thrown up by this step highlights the fact that there appears to be no straightforward match between the pattern and the signals. That is, texts may be constructed on a Problem-Solution pattern but not include any of the central lexical signals of the pattern. Of more particular interest to Scott is the fact that the words which can signal the pattern may occur without that part of the text being constructed as Problem-Solution; or they may have a purely local scope (e.g. the discussion of the problem takes up only a sentence or two) rather than serving to organise the text as a whole. This latter point raises the question of how readers (unlike the computer program) are able to identify whether a word such as problem is a marker of a text pattern or simply a local, incidental mention.

Overall, the relationship between keyness and signals of text structuring, which Scott sets out to explore, remains elusive. Like Berber Sardinha’s paper, Scott’s study raises more questions than it answers — but that is their function. One message that the present collection highlights is that we have increasingly sophisticated tools for exploring the behaviour of words and clauses in text but that we are only just beginning to address seriously the issue of using those tools to explore the nature of texts as meaningful entities in themselves. It is only by trying to use the tools for this purpose that we can gradually build up a picture of which tools work, which need adapting — and which throw up unexpected aspects of texts that are worth exploring.