8. Aggregation

Conceptual aggregation, which in the case of medical documents can be understood as the substitution of several expressions denoting certain entities by a single term or phrase encompassing these, cannot be done without knowledge of the discourse domain. Consider, for instance, the decision whether “sclerosis of the RCA, the LCA and the rami marginales” should be described as “global coronary sclerosis”. Although this is the most common case, conceptual aggregation does not occur only along the “part-of” relation. A “mitral regurgitation” and a “stenosis of the mitral valve” should adequately be described as “combined mitral valve disease”. Supporting this type of aggregation is a very difficult task and it is not yet realized satisfactorily in Suregen-2.

Conjunction reduction is much more a syntactical task and can be construed as merging phrases or clauses which
 

• are implicitly or explicitly linked with an “and”, and
• share a common expression.

It is important to understand that not every construct with “and” can be aggregated this way: Sometimes the “and” carries the (temporal) meaning of “and (then we did so and so)” or is used to denote causality “and (therefore it happened that)”. These connotations are much weaker or even lost in the aggregated forms. It may even be the case that the new aggregation carries another meaning, not present in the isolated expressions (that “Mick is married and Carly is married” is not equivalent to “Mick and Carly are married” is the standard example of such a case. It is therefore necessary that the user specifies (in the ToDescribe-form) whether Aggregate is to be called or not.