Custom SpaCy 3.0 models in Rasa
In this guide, we're going to bootstrap a spaCy 3.0 project and show you how you can integrate it with Rasa. We will train an entity detector and demonstrate how to use the new spaCy projects feature.…
Why Rasa uses Sparse Layers in Transformers
By Johannes Mosig and Vladimir Vlasov. Feed forward neural network layers are typically fully connected, or dense. But do we actually need to connect every input…
Exploring Semantic Map Embeddings / Part II
In this second part of our series on semantic maps, we show how to create them and see how they perform as featurizers for DIET.…
Exploring Semantic Map Embeddings / Part I
We explore a new sparse text embedding that has some interesting properties.…
Introducing Entity Roles and Groups
At a fundamental level, natural language understanding (NLU) does two things: it identifies the goal or meaning of the text and extracts key pieces of information…
10 Best Practices for Designing NLU Training Data
Whether you're starting from scratch or working with an existing data set, here's how to make sure your NLU training data results in accurate predictions and scales sustainably.…
Introducing DIET: state-of-the-art architecture that outperforms fine-tuning BERT and is 6X faster to train
With Rasa 1.8, our research team is releasing a new state-of-the-art lightweight, multitask transformer architecture for NLU: Dual Intent and Entity Transformer (DIET).…