NLP Day is a hands-on event, focused on practical battle-tested frameworks, from professional mentors who practice natural-language-processing for a living. Participants will bring their own laptop, and tackle common NLP tasks together. Our workshops assume you already know python and basic machine learning, and will guide you through the wonderful world of practical NLP.
Meet our team of natural language experts
Register, get swag, grab some coffee and snacks and do some mingling.
Recent years have seen a major jump in state-of-the-art results on various NLP tasks, with the introduction of powerful transformer-based deep neural networks trained on huge corpora. But when attempting to build a text classifier for our own custom domain, what does it all mean for us? In this workshop, I’ll walk you through building an effective text classifier using only a handful of labeled data points. We’ll label data using active learning and guided search, evaluate the performance of our model and our labels, use weak learners and data programming with the Snorkel package and employ state-of-the-art models (e.g. BERT) to our own data. We’ll discuss common pitfalls and eventually obtain a working, high quality text classifier in a matter of hours.
Deep learning methods have revolutionized the NLP field, breaking state-of-the-art benchmarks in classification, named-entity recognition and translation. The dark-side of deep learning, is the vast amount of labeled data required to train a model. This workshop would focus on classification and named-entity-recognition (NER) with deep learning methods on data that’s freely available on the internet. We would demonstrate training your-own classifier from scratch, and automatic formatting of text with sequence labeling techniques.
Bon Apetit!
spaCy had become the standard go-to library when practicing NLP, with lightning fast pre-trained models, and an extendable interface. In this workshop, we will demonstrate how we can take advantage of spaCy’s pipeline (specifically Named Entity Recognition and Dependency Parsing) capabilities to identify newly released products from internet data.
Without you it would not be possible