Evaluating the State-of-the-Art of End-to-End Natural Language Generation: The E2E NLG Challenge
arXiv:1901.07931 · doi:10.1016/j.csl.2019.06.009
Abstract
This paper provides a comprehensive analysis of the first shared task on End-to-End Natural Language Generation (NLG) and identifies avenues for future research based on the results. This shared task aimed to assess whether recent end-to-end NLG systems can generate more complex output by learning from datasets containing higher lexical richness, syntactic complexity and diverse discourse phenomena. Introducing novel automatic and human metrics, we compare 62 systems submitted by 17 institutions, covering a wide range of approaches, including machine learning architectures -- with the majority implementing sequence-to-sequence models (seq2seq) -- as well as systems based on grammatical rules and templates. Seq2seq-based systems have demonstrated a great potential for NLG in the challenge. We find that seq2seq systems generally score high in terms of word-overlap metrics and human evaluations of naturalness -- with the winning SLUG system (Juraska et al., 2018) being seq2seq-based. However, vanilla seq2seq models often fail to correctly express a given meaning representation if they lack a strong semantic control mechanism applied during decoding. Moreover, seq2seq models can be outperformed by hand-engineered systems in terms of overall quality, as well as complexity, length and diversity of outputs. This research has influenced, inspired and motivated a number of recent studies outwith the original competition, which we also summarise as part of this paper.
arXiv commentComputer Speech and Language, final accepted manuscript (in press)
References in corpus (26)
- Adam: A Method for Stochastic Optimization
- Neural Machine Translation by Jointly Learning to Align and Translate
- Sequence to Sequence Learning with Neural Networks
- Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
- Microsoft COCO Captions: Data Collection and Evaluation Server
- Pathologies of Neural Models Make Interpretations Difficult
- Pointer Networks
- Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems
- Learning to Search Better Than Your Teacher
- Stochastic Language Generation in Dialogue using Recurrent Neural Networks with Convolutional Sentence Reranking
- Attention-Based Recurrent Neural Network Models for Joint Intent Detection and Slot Filling
- Sequence-to-Sequence Generation for Spoken Dialogue via Deep Syntax Trees and Strings
- Challenges in Data-to-Document Generation
- A Context-aware Natural Language Generator for Dialogue Systems
- Three Generative, Lexicalised Models for Statistical Parsing
- Describing a Knowledge Base
- Natural Language Generation in Dialogue using Lexicalized and Delexicalized Data
- Referenceless Quality Estimation for Natural Language Generation
- Crowd-sourcing NLG Data: Pictures Elicit Better Data
- Learning Neural Templates for Text Generation
- Constrained Decoding for Neural NLG from Compositional Representations in Task-Oriented Dialogue
- Dual Supervised Learning for Natural Language Understanding and Generation
- Learning to Generate Wikipedia Summaries for Underserved Languages from Wikidata
- Investigating Linguistic Pattern Ordering in Hierarchical Natural Language Generation
- Natural Language Generation by Hierarchical Decoding with Linguistic Patterns
- Characterizing Variation in Crowd-Sourced Data for Training Neural Language Generators to Produce Stylistically Varied Outputs
Cited by in corpus (36)
- Survey on Evaluation Methods for Dialogue Systems
- Evaluation of Text Generation: A Survey
- The GEM Benchmark: Natural Language Generation, its Evaluation and Metrics
- C2Ideas: Supporting Creative Interior Color Design Ideation with Large Language Model
- HarmonyCut: Supporting Creative Chinese Paper-cutting Design with Form and Connotation Harmony
- Anticipating Safety Issues in E2E Conversational AI: Framework and Tooling
- Metaphoric Paraphrase Generation
- Learning to Reason for Text Generation from Scientific Tables
- Latent Space Factorisation and Manipulation via Matrix Subspace Projection
- Logical Natural Language Generation from Open-Domain Tables
- Neural Data-to-Text Generation via Jointly Learning the Segmentation and Correspondence
- Improved Natural Language Generation via Loss Truncation
- Constrained Decoding for Neural NLG from Compositional Representations in Task-Oriented Dialogue
- Data-to-Text Generation with Style Imitation
- RevCore: Review-augmented Conversational Recommendation
- Towards Faithfulness in Open Domain Table-to-text Generation from an Entity-centric View
- KGPT: Knowledge-Grounded Pre-Training for Data-to-Text Generation
- Logic2Text: High-Fidelity Natural Language Generation from Logical Forms
- Exploring Versatile Generative Language Model Via Parameter-Efficient Transfer Learning
- Towards transparency in NLP shared tasks
- CrossFit: A Few-shot Learning Challenge for Cross-task Generalization in NLP
- Partially-Aligned Data-to-Text Generation with Distant Supervision
- Sentence-Level Content Planning and Style Specification for Neural Text Generation
- Document Automation Architectures and Technologies: A Survey
- A Text Reassembling Approach to Natural Language Generation
- A Systematic Review of Open Datasets Used in Text-to-Image (T2I) Gen AI Model Safety
- Enhancing Task-Oriented Dialogues with Chitchat: a Comparative Study Based on Lexical Diversity and Divergence
- AGGGEN: Ordering and Aggregating while Generating
- DORB: Dynamically Optimizing Multiple Rewards with Bandits
- Machine Reading Comprehension using Case-based Reasoning
- LOGEN: Few-shot Logical Knowledge-Conditioned Text Generation with Self-training
- R2D2: Relational Text Decoding with Transformers
- Neural Path Hunter: Reducing Hallucination in Dialogue Systems via Path Grounding
- Data-QuestEval: A Referenceless Metric for Data-to-Text Semantic Evaluation
- Transforming Multi-Conditioned Generation from Meaning Representation
- Let Your Heart Speak in its Mother Tongue: Multilingual Captioning of Cardiac Signals