DTLEX: Ein Expertensystem zur Objektbestimmung auf der Basis natürlichsprachiger Entscheidungsbäume

We report on the development of an expert system for the determina-
tion of objects, whose identification is difficult. The system is
called DTLEX = "Decision Tree and natural Language oriented Expert
System". It implements the technique of keys formulated in natural
language used especially in biology. These keys can be interpreted as
decision trees. In addition and in parallel to the normal step by
step decision process, properties and attributes of the unknown
object can be used very similar to the methods used in information
retrieval in oder to get an object identification. This function
requires a preprocessing of the properties formulated in natural
language. For that purpose the properties and attributes were supp-
lied automatically with mechanically resolvable semantic classifica-
tions. The source text used is a large key for the determination of
approximately 3100 european mushrooms. The whole knowledge base of
DTLEX is represented by a text file containing among other things
more than 65000 semantically classified attributes. (Paper in German)