The dream of the early computer era was that of a more or less intelligent
electronic machine that understands its human counterpart and communicates
bidirectionally using their language. The history shows that expecting
computers to be able to “sit at the table” and converse with humans in a few
years was a naïve idea, and that the field of human-computer communication
had a long way to go.

For a computer it is not easy to “understand” the fuzzy, potentially
ambiguous natural language (e.g., English). In fact, the greatest portion,
by far, of the research work in the field of Natural Language Processing
(NLP) revolves around the analysis of language, rather than its production.
All the while, the generation of proper expressions in natural language is
an open problem. Applications that need to generate text on-the-fly based on
a variety of possible inputs are essential but not easy to implement.

The ultimate goal of a system that does Natural Language Generation is to
convey a meaning, simple or complex, through natural language. But how do we
describe and represent meaning in the first place? Here, we make use of the
logical forms from Discourse Representation Theory, which has been
successfully employed to represent and reason about abstract meaning in
a formal way. Logical forms are formal by definition, thus also directly
tractable with computational methods. This thesis presents a modular system
based on computational methods to generate natural language expressions from
these kind of input.