Sentence examples for sentence representations from inspiring English sources

The resulting abstract sentence representations are similar to the previous Example 1 in Sect.

Surface realisation can be viewed as the task of mapping abstract sentence representations to word strings in a target language.

In natural language generation (NLG) systems, surface realisation is known as the task of mapping abstract sentence representations to a surface form, that is, a sequence of words, punctuation symbols etc., to be delivered to the document presentation system [1].

The model makes a hard attention through the composition from sentence representation into document representation.

The system takes as an input an under-specified abstract sentence representation (cf. Sect. 3.2).

The work in [2], for instance, considers its own sentence representation in the form of labelled directed graphs.

The top-level sentence representation is the sentence clause, in this case representing a fixed order Agent-Action-Patient-Punctuation sentence template.

So first the model learns sentence representation where we can gain from both past and future input features thanks to Bidirectional Long Short-Term Memory Networks architecture.

The input to the system is a shallow, possibly under-specified sentence representation similar to the specification considered in the Surface Realisation Shared Task [12].

We described a 2-stage shallow surface realisation system for Brazilian Portuguese generation that takes as an input an abstract sentence representation and over-generates multiple candidate sentences with the aid of a small set of language-dependent linearisation rules.

Each test sentence was converted into an abstract sentence representation by performing two tasks: (1) every content word was replaced by its lemma and (2) the constituent ordering of NP and VP modifiers was randomised.