A REFERENCE ARCHITECTURE FOR GENERATION SYSTEMS GUIDE

Posted by Anonymous 0 komentar

Download free A Reference Architecture for Generation Systems Guide pdf. This paper describes an attempt to specify a reference architecture for natural language generation systems. The ﬁeld of Natural Language Generation (NLG) is characterised by a variety of theoretical approaches, which has the eﬀect of fragmenting the community and reducing the bandwidth of communication. The technology is mature enough for commercial applications to be close, and yet it is relatively rare for NLG software and data to be reused between diﬀerent researchers. The rags (Reference Architecture for Generation Systems) project set out to exploit the implicit agreement that does exist in the ﬁeld and focus discussion on the genuine points of theoretical disagreement with the aid of a “standard” vocabulary and reference point. Thus rags aimed to make it easier to:

create reusable data resources (e.g. representative inputs to algorithms, corpora of generated texts together with their underlying representations);
communicate data between NLG program modules written at diﬀerent times and with diﬀerent assumptions, and hence allow reuse of such modules;
allow modules (or at least their inputs and outputs) to be deﬁned in a relatively formal way, so that their scope and limitations may be better understood.

In this way, applications-oriented systems development should be facilitated, standard interfaces and datasets can emerge, ideas and software can be more easily reused, researchers can realistically specialise in particular areas and comparative evaluation of systems and components may become possible.

here