TNT-NLG , System 2 : Data Repetition and Meaning Representation Manipulation to Improve Neural Generation

End-to-End (E2E) neural models that learn and generate natural language sentence realizations in one step have recently received a great deal of interest from the natural language generation (NLG) community. In this paper, we present “TNT-NLG” System 2, our second system submission in the E2E NLG challenge, which focuses on generating coherent natural language realizations from meaning representations (MRs) in the restaurant domain. We tackle the problem of improving the output of a neural generator based on the open-source baseline model from Dusek et al. (2016) by vastly expanding the training data size by repetition of instances in training, and permutation of the MR. We see that simple modifications allow for increases in performance by providing the generator with a much larger sample of data for learning. Our system is evaluated using quantitative metrics and qualitative human evaluation, and scores competitively in the challenge.

[1]  Andreas Vlachos,et al.  Imitation learning for language generation from unaligned data , 2016, COLING.

[2]  Ondrej Dusek,et al.  Sequence-to-Sequence Generation for Spoken Dialogue via Deep Syntax Trees and Strings , 2016, ACL.

[3]  Dilek Z. Hakkani-Tür,et al.  To Plan or Not to Plan? Sequence to sequence generation for language generation in dialogue systems , 2017 .

[4]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[5]  Hiroaki Kitano,et al.  RoboCup: Robot World Cup , 1998, CROS.

[6]  Victor Zue,et al.  Experiments in Evaluating Interactive Spoken Language Systems , 1992, HLT.

[7]  Ehud Reiter,et al.  Book Reviews: Building Natural Language Generation Systems , 2000, CL.

[8]  Michael White,et al.  Enhancing the Expression of Contrast in the SPaRKy Restaurant Corpus , 2013, ENLG.

[9]  Matthew R. Walter,et al.  What to talk about and how? Selective Generation using LSTMs with Coarse-to-Fine Alignment , 2015, NAACL.

[10]  David Vandyke,et al.  Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems , 2015, EMNLP.

[11]  Oliver Lemon,et al.  Crowd-sourcing NLG Data: Pictures Elicit Better Data. , 2016, INLG.

[12]  Yoshua Bengio,et al.  Neural Machine Translation by Jointly Learning to Align and Translate , 2014, ICLR.

[13]  Jianfeng Gao,et al.  A Neural Network Approach to Context-Sensitive Generation of Conversational Responses , 2015, NAACL.

[14]  Verena Rieser,et al.  The E2E Dataset: New Challenges For End-to-End Generation , 2017, SIGDIAL Conference.

[15]  Verena Rieser,et al.  Findings of the E2E NLG Challenge , 2018, INLG.

[16]  Marilyn A. Walker,et al.  A Personality-based Framework for Utterance Generation in Dialogue Applications , 2008, AAAI Spring Symposium: Emotion, Personality, and Social Behavior.

[17]  Milica Gasic,et al.  Training and Evaluation of the HIS POMDP Dialogue System in Noise , 2008, SIGDIAL Workshop.

[18]  Frédéric Béchet,et al.  The French MEDIA/EVALDA Project: the Evaluation of the Understanding Capability of Spoken Language Dialogue Systems , 2004, LREC.

[19]  Johanna D. Moore,et al.  Fish or Fowl:A Wizard of Oz Evaluation of Dialogue Strategies in the Restaurant Domain , 2002, LREC.

[20]  Larry P. Heck,et al.  Contextual LSTM (CLSTM) models for Large scale NLP tasks , 2016, ArXiv.

[21]  Matt Post,et al.  Efficient Elicitation of Annotations for Human Evaluation of Machine Translation , 2014, WMT@ACL.

[22]  Milica Gasic,et al.  Phrase-Based Statistical Language Generation Using Graphical Models and Active Learning , 2010, ACL.

[23]  Marilyn A. Walker,et al.  Trainable Sentence Planning for Complex Information Presentations in Spoken Dialog Systems , 2004, ACL.