Adapting text to data in documents through a natural language processor

摘要

This paper presents an approach to the problem of generating documents from schemas and variable data so that the schematic contents are made linguistically consistent with the injected data.A key aspect of the proposed approach consists in performing most of the required linguistic analysis during the schema definition phase instead of during runtime document production. This is achieved by abstracting at schema definition time the main linguistic properties of the data which will be injected.Another important characteristic of the approach consists in the fact that not only variable data can be injected in the schematic contents but also a whole document; in this case the resulting document is linguistically adjusted, as a whole, resolving references between the outer and the inner subdocuments.