Machine learning concepts

This page includes guides and resources on the machine learning concepts behind Communications Mining, and are listed in the following table:

Guide	Description
How machines learn to understand words: a guide to embeddings in NLP	How Communications Mining uses Transformer-based embeddings to represent text semantically and power its machine learning models.
Prompt-based learning with Transformers	How prompt-based learning with Transformer models improves natural language processing tasks.
Efficient Transformers II: knowledge distillation & fine-tuning	How knowledge distillation and fine-tuning make Transformer-based NLP models more efficient.
Efficient Transformers I: attention mechanisms	How attention mechanisms make Transformer-based NLP models more efficient.
Deep hierarchical unsupervised intent modelling: getting value without training data	How deep hierarchical unsupervised intent modelling extracts value from communications without training data.
Fixing annotating bias with Communications Mining™	What causes annotation bias in machine learning models and how to remediate it.
Active learning: better ML models in less time	How active learning reduces the annotation effort needed to train accurate machine learning models.
It's all in the numbers: assessing model performance with metrics	How to interpret the performance metrics used to evaluate machine learning models.
Why model validation is important	Why model validation matters and the risks of deploying an unvalidated model.
Comparing Communications Mining™ and Google AutoML for conversational data intelligence	How Communications Mining compares with Google AutoML for NLP-driven process automation and conversational data intelligence.