Eder Miranda de Novais

Improved text generation using n-gram statistics

In Natural Language Generation (NLG) systems, a general-purpose surface realisation module will usually require the underlying application to provide highly detailed input knowledge about the target sentence. As an attempt to reduce some of this complexity, in this paper we follow a traditional approach to NLG and present a number of experiments involving the use of n-gram language models as an aid to an otherwise rule-based text generation approach. By freeing the application from the burden of providing a linguistically- rich input specification, and also by taking some of the generation decisions away from the surface realisation module, we expect to make NLG techniques accessible to a wider range of potential applications.

A Testbed for Portuguese Natural Language Generation

We present a data-text aligned corpus for Brazilian Portuguese Natural Language Generation (NLG) called SINotas, which we believe to be the first of its kind. SINotas provides a testbed for research on various aspects of trainable, corpus-based NLG, and it is the basis of a simple NLG application under development in the education domain.

Text-to-text surface realisation using dependency-tree replacement

Surface realisation - the task of producing word strings from non-linguistic input data - has been the focus of a great deal of research in the field of data-to-text Natural Language Generation (NLG). In this work we discuss an alternative approach to surface realisation, in which we borrow NLG techniques from the sister field of text-to-text generation to implement text generation based on examples in natural language. Our approach is suitable to simpler applications that are not linguistically-oriented by design, and which may be able to provide only minimal input knowledge to the NLG module.