Introduction

With the advent of large-scale electronic text collections, today’s corpus linguists have a vast amount of potentially interesting data at their disposal to study a wide range of questions. Often, this data is annotated in a number of ways. For example, many corpora include various types of meta-textual information that make it possible to investigate differences in language use as a reflection of factors such as mode (speech vs. writing), text type (e.g. academic prose vs. fiction), manner of interaction (e.g. monologue vs. dialogue) and the socio-demographic characteristics of language users (e.g. sex, age, social class of author/speaker).

Apart from these text or speaker-level annotations, modern corpora often also contain some level of linguistic annotation. Perhaps the most common type is the grammatical annotation of each word in the corpus with a so-called part-of-speech tag (POS-tag). With the help of a suitable corpus retrieval program, these POS-tags make it possible to search for grammatical structures by defining sequences of tags. For example, a researcher interested in the phenomenon of intensification will find most relevant instances in the corpus by way of a tag sequence that retrieves all cases where an adverb immediately precedes an adjective (e.g. very good, especially important, exceedingly difficult). Although this sequence in fact also retrieves a number of irrelevant hits (e.g. currently available), this method is a more effective procedure than trying to compile a list of relevant lexical items (i.e. all potential intensifiers) from scratch and then searching for each of them individually. Depending on the feature-set of the corpus tool at hand, the different levels of annotation available in the corpus can easily be combined in searches, thus enabling the description of grammatical variation across different language uses and settings with comparatively little effort.