We describe a corpus of numerical expressions, developed as part of the NUMGEN project. The corpus contains newspaper articles and scientific papers in which exactly the same numerical facts are presented many times (both within and across texts). Some annotations of numerical facts are original: for example, numbers are automatically classified as round or non-round by an algorithm derived from Jansen and Pollmann (2001); also, numerical hedges such as ‘about’ or ‘a little under’ are marked up and classified semantically using arithmetical relations. Through explicit alignment of phrases describing the same fact, the corpus can support research on the influence of various contextual factors (e.g., document position, intended readership) on the way in which numerical facts are expressed. As an example we present results from an investigation showing that when a fact is mentioned more than once in a text, there is a clear tendency for precision to increase from first to subsequent mentions, and for mathematical level either to remain constant or to increase.

This paper describes a study of the levels at which different rhetorical relations occur in rhetorical structure trees. In a previous empirical study (Williams and Reiter, 2003) of the RST-DT (Rhetorical Structure Theory Discourse Treebank) Corpus (Carlson et al., 2003), we noticed that certain rhetorical relations tended to occur more frequently at higher levels in a rhetorical structure tree, whereas others seemed to occur more often at lower levels. The present study takes a closer look at the data, partly to test this observation, and partly to investigate related issues such as the relative complexity of satellite and nucleus for each type of relation. One practical application of this investigation would be to guide discourse planning in Natural Language Generation (NLG), so that it reflects more accurately the structures found in documents written by human authors. We present our preliminary findings and discuss their relevance for discourse planning.