Data-to-text generation maps numeric and/or symbolic representations of information to language that expresses the information in some way, for example creating summaries, reports and descriptions from databases and other sources of non-verbal information.
The number of potential applications of data-to-text NLG technology is vast: most companies and organisations regularly turn some form of nonverbal information (e.g. account information, performance results) into letters, reports, manuals, etc. NLG technology can help economise text-production processes, but it can also make information available in verbal form that would otherwise be inaccessible or more time-consuming to process.
NLG researchers have looked at a wide range of applications which broadly fit under two headings:
1. Economising text production: serial letters for business and health providers; descriptions of museum exhibits, stock market movements, route directions; instructional texts such as cooking recipes, technical manuals, patient information leaflets; prognostic reports such as weather forecasts and pollen forecasts;
2. Making nonverbal information more available:
(a) increasing accessibility to non-experts—explanations of proofs and complex phenomena, presenting information in expert systems;
(b) speeding up the rate at which information can be processed— diagnostic reports such as medical reports, fault reports, error messages, air pollution reports;
(c) providing a modality that might otherwise be wasted—e.g. where visual modalities are already busy, additional, nonverbal information can be converted into verbal form and provided over an audio channel.