PILLS: Multilingual generation of medical information documents with overlapping content
Nadjet Bouayad-Agha (ITRI, University of Brighton, Lewes Road, Brighton BN2 4GJ, UK)
Richard Power (ITRI, University of Brighton, Lewes Road, Brighton BN2 4GJ, UK)
Donia Scott (ITRI, University of Brighton, Lewes Road, Brighton BN2 4GJ, UK)
Anja Belz (ITRI, University of Brighton, Lewes Road, Brighton BN2 4GJ, UK)
WO24: Applications Based On Written LRs
In the pharmaceutical industry, products have to be described by a range of document types with overlapping content. Moreover, much of this documentation has to be produced in many languages. This situation is commonplace in many commercial domains, and leads to well-known problems in maintaining a set of related documents and their translations. We describe a potential solution explored in the PILLS project. All relevant knowledge about a product is entered only once, through a natural-language interface to a knowledge base. From this `master model', specialised models for arange of document types are derived automatically; from each specialised model, documents are generated automatically in all supported languages. As an illustration of this approach, the PILLS demonstrator generates three medical document types in English, German and French.
Content determination, Multilingual authoring, Natural language generation, Pharmaceutical and healthcare publishing