Generation of texts for information graphics

We describe SelTex a text generation system for producing short texts and captions to accompany information graphics that are generated according to the writer's intentions. SelTex uses rules that were extracted from a corpus study of more than 400 text excerpts. This corpus study shows that text and graphics play complementary roles in transmitting information from the writer to the reader. We then derive some observations for the automatic generation of texts associated with graphics many of which were implemented in SelTex. For the past few years, we have studied the automatic generation of graphics from statistical data in the context of the PostGraphe system (5, 6). PostGraphe is given data in tabular form as might be found in a spreadsheet; also input is a declaration of the types of values in the columns of the table. The user then indicates the intentions to be conveyed in the graphics (e.g. compare two variables or show the evolution of a set of variables) and the system generates a report in L ATEX with the appropriate PostScript graphic files. PostGraphe also generates an accompanying text for only a few simple text schemas. Although PostGraphe has shown the potential of graphic generation, its text generation capabilities are well below its level of graphic competence. So we decided to redesign the text generation module to better capture the subtle nature of the interaction between text and graphics. Before designing SelTex, the new text module to be integrated in PostGraphe, we did a corpus study of more than 400 texts associated with graphics. The preliminary findings were presented in (3). They will be reviewed here briefly before describing the text generation rules that were deduced from this corpus study. We will focus on the selection criteria for relevant facts that are expressed in the text that accompanies a graphic.