Sentence Packaging in Text Generation from Semantic Graphs as a Community Detection Problem

Abstract

An increasing amount of research tackles the challenge of text generation from abstract ontological or semantic structures, which are in their very nature potentially large connected graphs. These graphs must be "packaged" into sentence-wise subgraphs. We interpret the problem of sentence packaging as a community detection problem with post optimization. Experiments on the texts of the Verb-Net/FrameNet structure annotated-Penn Treebank, which have been converted into graphs by a coreference merge using Stanford CoreNLP, show a high F 1-score of 0.738.

MindSpaces: Art-driven Adaptive Outdoors and Indoors Design

“…The text generation service takes structured input in terms of ontological structures, which can originate from textual analysis of the previously provided material or from non-textual sources such as the resulting structures from processing of EEG or visual signals (i.e., emotions and the visual behavior events/activities), and verbalizes the information in the language of the preference of the reader. Due to the stratified organization of its lexical and grammatical resources being elaborated within undergoing continuous research (Shvets, Mille, & Wanner, 2018), (Mille, Belz, Bohnet, & Wanner, 2018), multilingual sentence generation provides a possibility to start from input structures of different levels of abstraction (including ontological representations). Due to their generalization-oriented modularization, the text generator can be rather easily adapted to new languages and new genres (and thus styles).…”