Reuse/adaptation in the TSM
 Reuse of INSTRUCTIONS-region
nIn building regions for other text types within the same domain
nTransferability to different domains
 Relatively domain/language
 independent sentence planner (SPL)
 Multilinguality in the TSM
nSharing principal decisions in text planning
nControl over language-specific realization in SPLs
• Reuse of INSTRUCTIONS-region
•In building regions for other text types within the same domain:
The development of the TSM started with the goal to plan instructional texts. To this end, we developed a region called the INSTRUCTIONS-region, which creates a text plan that plans (linearizes) the content specified in an A-box. The region implements a number of basic strategies for planning various levels of aggregation, sequence marking, etc. We have reused the same strategies, and sometimes even the same basic setup of the region, to implement regions for planning different text types.
•Transferability to different domains
As said, the strategies are independent, and we stipulate that the region can be immediately reused for other task-oriented domains, where the difference is in content rather than in the organization of the content. If the content is arranged differently as well (i.e. the domain model would contain different configurational concepts), then it depends on the mapping or relation that one could establish between the configurational concepts defined in the AGILE DM and the DM of the new domain.
• Relatively domain/language independent sentence planner (SPL): The basic architecture of the sentence planner is such that its decision making is minimal – it interprets the structure indicated in the text plan (how to aggregate content into sentence-sized chunks) rather than deciding this by itself. The sentence planner is easy to configure/extend to cover more constructions (e.g. more RST relations). It is of course “relatively” domain/language independent in that particular idiosyncrasies of the project and its languages have made their way into the sentence planner – although this is, we hasten to say, minimal. The strategy that the sentence planner follows, being a read-out of the text plan, creating content for the leaves, inserting traces in the set of leave-content-SPLs, and then we recursing back over the textplan, interpreting the traces put into the text and creating complex SPLs if needed, is generally applicable.
• Multilinguality in the TSM
•Sharing principle decisions in text planning: The text planning regions are entirely multilingual.
•Control over language-specific realization in SPLs: Where the language-specific items come in are the systems where we control language-specific realizations of particular text plan elements – for example, titles might be realised differently across languages. We can exert that kind of control in the text planner already.