Abstract

Traditional research in NLG focuses on building better models and assessing their performance using clean, preprocessed and curated datasets, as well as standard automatic evaluation metrics. From a scientific point-of-view, this provides a controlled environment where different models can be compared and robust conclusions can be made. However, these controlled settings can drastically deviate from scenarios that happen when deploying systems in the real world. In this talk, I will focus on what happens before data is fed into NLG systems and what happens after we generate outputs. For the first part, I will focus on addressing heterogeneous data sources using tools from graph theory and deep learning. In the second part, I will talk about how to improve decision making from generated texts through Bayesian techniques, using Machine Translation post-editing as a test case.

Bio: Daniel is a Lecturer at The University of Melbourne. His main research topic is Natural Language Generation, with a focus on Machine Translation. He is particularly interested in using tools from Machine Learning, Theoretical Computer Science and Statistics to address challenges in NLG that go beyond the usual input-output pipeline. He obtained a PhD from The University of Sheffield, United Kingdom, and his thesis on using Gaussian Processes for NLP applications received a Best Thesis Award from the European Association for Machine Translation. Daniel is also an advocate for queer and LGBT + visibility in STEM , in particular within NLP and Machine Learning. He is currently a board member of the Widening NLP initiative (www.winlp.org), which foster inclusivity from underrepresented groups in NLP . His personal webpage can be found at https://beckdaniel.wordpress.com and he tweets at https://twitter.com/beck_daniel