We enjoyed another fascinating AI in Action Virtual meetup this past Wednesday evening as we were joined by three excellent speakers to share their ground-breaking work at the second of our NLP focused series.
Thomas Vrancken, Machine Learning Engineer at ML6 kicked things off by telling us about how they are designing and improving real-world NLP for production systems. He specifically focused on the insights they obtained during the development and go-live of a recent NLP-powered client application which is a is a search engine based on document embeddings.
Thomas talked us through what worked and what didn't. He also explored how to host their Machine Learning models on google cloud, integrating the search engine with elastic search and how they use kubeflow pipelines to automate their workflows.
Next up Procheta Sen from the ADAPT Centre DCU introduced us to her learnings in Socially Responsible AI: Cognitive Bias-Aware Multi-Objective Learning. The prevalent existence of cognitive biases in large volumes of historical data can pose a threat of being manifested as unethical and seemingly inhumane predictions as outputs of AI systems trained on such data.
To alleviate this problem, Procheta and her fellow students propose a bias-aware multi-objective learning framework that given a set of identity attributes (e.g. gender, ethnicity etc.) and a subset of sensitive categories of the possible classes of prediction outputs, learns to reduce the frequency of predicting certain combinations of them.
Last but not least Alexander O'Connor, Lead Data Scientist at Autodesk gave us an industry perspective into the latest trends in NLP. He specifically touched on his experience with Transformers, which have been the dominant technology in natural language processing for more than a year now.
The most exciting thing is that technology is growing capability month by month. Alex shared his insights into the astonishing progress of Transformers, along with the unprecedented availability of code and data, make it a compelling tool for industrial applications.
You can now watch the meetup in full below.