In lieu of an abstract, here is a brief excerpt of the content:

Reviewed by:
  • Foundations of computational linguistics: Man-machine communication in natural language by Roland Hausser
  • Michael A. Covington
Foundations of computational linguistics: Man-machine communication in natural language. By Roland Hausser. Berlin: Springer, 1999. Pp. xii, 534.

‘Today’s situation in computational linguistics resembles the development of mechanical flight before 1903. For hundreds of years humans have observed . . . birds in order to understand how they fly. Their goal was to become airborne in a similar manner.

It turned out, however, that flapping wings did not work for humans. This was taken by some to declare human flight impossible in principle . . .

Today human air travel is commonplace. Furthermore, we now know that a sparrow remains airborne in terms of the same aerodynamic principles as a jumbo jet’ (4).

Thus Hausser begins his survey of computational linguistics which, although presented (and usable) as a textbook, is definitely Grundlagen in the German sense—everything is developed from first principles, and rigor is more important than comprehensiveness.

H’s key claim is that successful natural language processing (NLP) requires the right level of abstraction, somewhere between mathematics and psycholinguistics, just as airplane design required finding the right level between pure physics and ornithology. Thus H appreciates traditional grammar (‘because of its wealth of concrete data’, 17) and prefers surfaceoriented grammatical formalisms, especially categorical grammar, which he uses extensively. His four descriptive principles are summed up as SLIM, i.e.:

Surface compositionality (no inaudibilia, no transformations, no extralinguistic Gricean maxims or the like);

Linearity in time (avoidance of backtracking);

Internal ontology (modeling the speaker’s and hearer’s cognition, not the outside world); and

Matching of cognitive structures as the basis of reference.

I suspect this book’s main audience will consist, not of beginners, but of advanced workers who want to see familiar ideas developed in a more rigorous way. Formal pragmatics is a special strength of this book.

Michael A. Covington
University of Georgia
...

pdf

Share