Abstract: Deep learning finds patterns in input data with a sophistication and scale that other methods can’t achieve, yet often requires large labeled data sets that can be difficult to create for many applications in language understanding. Solving this issue is critical for deploying the real world where you are likely to have limited but specific datasets to create customized models for an application. Thus many people in NLP have relied on more generalized industry data, or spent months manually building and tuning models for their data. But recently, adding structured information such as knowledge graphs into deep learning models has proven to outperform pure end-to-end training for word embeddings, especially on smaller data sets. This method can be put into production without sacrificing scalability and accuracy.
In this session, we'll discuss how adding the digital intuition of a common sense knowledge base frees machine learning algorithms from big data requirements, accelerates time-to-value, and automates the human-supervised tuning work, so you can take advantage of machine learning using your specific data, not generic Big Data. She'll introduce algorithms such as retrofitting and discuss the implications of these advances for practical text analytics understanding. If we have time, we'll talk about extending these models to build higher order cognitive systems.
Bio: Dr. Catherine Havasi is Chief Strategy Officer and co-founder of Luminoso, an AI-based natural
language understanding company in Cambridge, MA. Previous to Luminoso, she directed the Digital Intuition group at MIT's Media Lab working on word embeddings, transfer learning, and language understanding. In the late 90s, she co-founded the Common Sense Computing Initiative, or ConceptNet, the first crowd-sourced project for artificial intelligence. ConceptNet has played a role in thousands of AI projects and will be turning 20 next year.