Many neural NLG systems “hallucinate” non-existent or incorrect content. This is a major problem, since such hallucination is unacceptable in many (most?) NLG use cases. Also BLEU and related metrics do not detect hallucination well, so researchers who rely on such metrics may be misled about the quality of their system.
From a commercial perspective, I think NLG is currently most successful in financial reporting. Although of course there are many great NLG applications in other sectors!
Some thoughts on how to vary words in NLG text. This is aimed at practioners who are building NLG systems, not researchers.
An approach I often take to NLG and indeed AI is to try to understand underying linguistic, NLG, and AI issues, and then to look for simple solutions to these issues.
Some thoughts on language grounding, especially choosing words to express data, and how this depends on context.
Unfortunately I suspect many researchers make their results looks better by using poor baselines. I give some thoughts on this, based on a recent discussion with a PhD student.
Some thoughts about when I feel comfortable being a coauthor on a paper, expressed as a letter to someone who put me on a paper as a co-author without asking me frst,
Some musings on principled and theoretically sound techniques for automatically evaluating NLG systems.
My advice on how to perform a high-quality validation study, which assesses whether a metric (such as BLEU) correlates well with human evaluations.
BLEU works much better for MT systems and NLG systems. In this blog I present some speculations as to why this is the case.
Ehud Reiter's Blog 