Abstract
Computational semantics is the branch of computational linguistics that is concerned with the development of methods for processing meaning information. Because a computer system that analyzes natural language must be able to deal with arbitrary real-world sentences, computational semantics faces a number of specific challenges related to the coverage of semantic construction procedures, the efficient resolution of ambiguities, and the ability to compute inferences. After initial successes with logic-based methods, the mainstream paradigm in computational semantics today is to let the computer automatically learn from corpora. In this article, we present both approaches, compare them, and discuss some recent initiatives for combining the two.