Each month the Natural Language Processing meetup brings together researchers, industry professionals, and students to discuss of state-of-the-art NLP research and cutting-edge industry applications. The most recent meetup was held on Monday 17th July. Hosted by Workday at their offices in Smithfield, it featured an all-female panel with Dr. Sheila Castilho, Dr. Carla Parra Escartín, and Dr. Teresa Lynn, all of whom are postdoctoral researchers at the ADAPT Centre in DCU. These panellists helped to discuss the focus and theme of this month’s meetup; learning to evaluate models from a human perspective, which were deemed vital in achieving application of machine learning processes to different languages.
Dr. Castilho kicked the event off with an overview of human evaluation through a presentation entitled “Human Evaluation – Why do we need it?”, in which she discussed the importance of evaluation in enabling progress in the development of NLP systems. Dr. Castilho holds a Master’s Degree in NLP and a PhD from DCU. Currently, she is a postdoctoral researcher at the TraMOOC project at the ADAPT Centre, focusing on machine and human evaluation of automatically translated subtitles and user-generated content.
Dr. Parra Escartín continued the discussion with the presentation of a case study of human-computer interaction for machine translation post-editing. Dr. Parra Escartín’s talk, titled “Living on the edge: productivity gain thresholds in machine translation evaluation metrics”, centred around a project involving 10 professional translators. Dr. Parra Escartín obtained her PhD in Computational Linguistics at the University of Bergen (Norway) and is currently the Marie Sklodowska-Curie fellow within the EDGE programme at the ADAPT Centre in DCU.
Finally, Dr. Lynn presented the meetup’s attendees with the Universal Dependencies Project, which enables cross-lingual parsing by capturing linguistic universals across languages. Dr. Lynn is a postdoctoral researcher at the ADAPT Centre, having completed her PhD through a cotutelle agreement between Macquarie University Sydney (Australia) and DCU. She built the first dependency treebank for Irish and trained the first statistical dependency parsing models for Irish. The Universal Dependencies Project can be viewed at the following link: http://universaldependencies.org/
ADAPT is delighted to have been represented by three of its leading experts in this field at the last NLP meetup, and looks forward to future collaborations with this group.
Share this article: