Data-Driven Broad-Coverage Grammars for Opinionated Natural Language Generation (ONLG)

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1331–1341 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1122 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1331–1341 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1122 Data-Driven Broad-Coverage Grammars for Opinionated Natural Language Generation (ONLG) Tomer Cagan School of Computer Science The Interdisciplinary Center Herzeliya, Israel cagan.tomer@idc.ac.il Stefan L. Frank Centre for Language Studies Radboud University Nijmegen, The Netherlands s.frank@let.ru.nl Reut Tsarfaty Mathematics and Computer Science The Open University of Israel Ra’anana, Israel reutts@openu.ac.il Abstract Opinionated natural language generation (ONLG) is a new, challenging, NLG task in which we aim to automatically gener- ate human-like, subjective, responses to opinionated articles online. We present a data-driven architecture for ONLG that generates subjective responses triggered by users’ agendas, based on automatically acquired wide-coverage generative gram- mars. We compare three types of gram- matical representations that we design for ONLG. The grammars interleave different layers of linguistic information, and are in- duced from a new, enriched dataset we de- veloped. Our evaluation shows that gener- ation with Relational-Realizational (Tsar- faty and Sima’an, 2008) inspired grammar gets better language model scores than lexicalized grammars `a la Collins (2003), and that the latter gets better human- evaluation scores. We also show that con- ditioning the generation on topic models makes generated responses more relevant to the document content. 1 Introduction Interaction in social media has become increas- ingly prevalent nowadays. It fundamentally changes the way businesses and consumers behave (Qualman, 2012), it is instrumental to the success of individuals and businesses (Haenlein and Ka- plan, 2009) and it also affects political regimes (Howard et al., 2011; Lamer, 2012). In particu- lar, automatic interaction in natural language in social media is now a common theme, as seen in the rapid popularization of chat applications, chat-bots, and “smart agents” aiming to conduct human-like interactions in natural language. So far, generation of human-like interaction in general has been addressed mostly commercially, where there is a movement towards online re- sponse automation (Owyang, 2012; Mah, 2012), and movement away from script-based interaction towards interactive chat bots (Mori et al., 2003; Feng et al., 2006). These efforts provide an au- tomated one-size-fits-all type of interaction, with no particular expression of particular sentiments, topics, or opinions. In academia, work on gen- erating human-like interaction focused so far on generating responses to tweets (Ritter et al., 2011; Hasegawa et al., 2013) or taking turns in short di- alogs (Li et al., 2017). However, the architectures assumed in these studies implement sequence to sequence (seq2seq) mappings, which do not take into account topics, sentiments or agendas of the intended responders. Many real-world tasks and applications would benefit from automatic interaction that is gener- ated intendedly based on a certain user profile or agenda. For instance, this can help promoting a political candidate or a social idea in social me- dia, aiding people forming …

Content truncated to 500 words