Generating Paraphrases from DBPedia using Deep Learning

Abstract

Recent deep learning approaches to Natural Language Generation mostly rely on sequence-to-sequence models. In these approaches, the input is treated as a sequence whereas in most cases, input to generation usually is either a tree or a graph. In this paper, we describe an experiment showing how enriching a sequential input with structural information improves results and help support the generation of paraphrases.

Point at the Triple: Generation of Text Summaries from Knowledge Base Triples

“…There is a large body of literature that uses the encoder-decoder framework from machine translation (Cho et al, 2014;Sutskever et al, 2014;Bahdanau et al, 2014) for NLG (Sleimi and Gardent, 2016;Gardent et al, 2017;Chisholm et al, 2017;Mei et al, 2016;Lebret et al, 2016;Wiseman et al, 2017;Vougiouklis et al, 2018a;Liu et al, 2018;Li and Wan, 2018;Gehrmann et al, 2018;Yeh et al, 2018). The decoder, typically a multi-gated Recurrent Neural Network (RNN), formed of either Long Short-Term Memory cells (Hochreiter and Schmidhuber, 1997) or Gated Recurrent Units (Cho et al, 2014), is conditioned on a set of structured records and acts as a language model.…”

“…The decoder, typically a multi-gated Recurrent Neural Network (RNN), formed of either Long Short-Term Memory cells (Hochreiter and Schmidhuber, 1997) or Gated Recurrent Units (Cho et al, 2014), is conditioned on a set of structured records and acts as a language model. Adaptation of such systems have shown great potential at tackling various aspects of triples-to-text tasks ranging from microplanning by Gardent et al (2017) to generation of paraphrases by Sleimi and Gardent (2016). Pointer-generator networks have been brought up recently by Gu et al (2016) and See et al (2017) as alterations of the original pointer architecture proposed by Vinyals et al (2015).…”

A Master-Apprentice Approach to Automatic Creation of Culturally Satirical Movie Titles

“…Similar sequence-to-sequence based approaches have been used in the past for text paraphrasing task (Brad and Rebedea, 2017;Sleimi and Gardent, 2016), which shares its similarities with the task we are set to solve. This gives us a reason to believe that sequence-to-sequence approach is a viable way of implementing the apprentice.…”

BENGAL: An Automatic Benchmark Generator for Entity Recognition and Linking

“…This attention comes from the great number of published works such as (Cimiano et al, 2013;Duma and Klein, 2013;Ell and Harth, 2014;Biran and McKeown, 2015) which used RDF as an input data and achieved promising results. Moreover, the works published in the WebNLG (Colin et al, 2016) challenge, which used deep learning techniques such as (Sleimi and Gardent, 2016;, also contributed to this interest. RDF has also been showing promising benefits to the generation of benchmarks for evaluating NLG systems, e.g., (Gardent et al, 2017;Mohammed et al, 2016;Schwitter et al, 2004;Hewlett et al, 2005;Sun and Mellish, 2006).…”