Corpus tools and methods, today and tomorrow: Incorporating linguists' manual annotations

Today’s corpus tools offer the user a wide range of features that greatly facilitatethe linguistic analysis of large amounts of authentic language data (e.g. frequencydistributions, collocations, keywords, etc.). However, these tools typically fail toaddress the fundamental need of the linguist to add interpretive information to aconcordance or query result, by coding individual concordance lines forstructural, functional, discoursal, and other features in a flexible way. The abilityto add such qualitative data is indispensable to a fuller understanding of thephenomenon under investigation as it allows the linguist to produce morerigorous descriptions—and theories—about language in use.Our article has two aims: first, to assess the merits and drawbacks of existingsolutions, by surveying what can be achieved using state-of-the-art corpus toolsand generic database software; second, we draw up a set of desiderata andrecommendations for the incorporation of flexible encoding features into futurecorpus tools. We describe an initial step in this direction, with a recent enhancementto the BNCweb corpus analysis software. More generally, we hope oursuggestions will lead to linguists and software developers working together moreclosely to ensure that the needs of the former are provided for by the availabletechnology