Is the computer capable of recognizing different sentence types in a linguistic corpus such as COSMAS II (Mannheim), which has not been previously treated by a tagger or a parser? The answer is in fact no. However, in the present article it is shown that under certain circumstances an automatic distinction is nevertheless possible. Making use of a procedure that we have called Anfragezuspitzung (literally: making a query pointed; encirclement of a grammatical phenomenon by a combination of several specific queries), and taking as a starting point philological prior knowledge that has been gathered "by hand", it proves to be perfectly possible to arrive at a satisfactory result. With the example of sentence types in German, in particular the distinction between interrogative and exclamatory sentences, we demonstrate in this article that such a distinction can be carried out automatically with a high degree of accuracy, e.g. the distinction between War das eine gute Idee? (Was this a good idea?) and War das eine gute Idee! (What a good idea this was!).