Skip to main content

Skills demonstration

Instruction and application
In Progress

Using SpaCy for NER demo​

In this demo, we’ll customise SpaCy’s NER models to recognise industry-specific entities, enhancing their relevance for government and various sectors.

This hands-on session demonstrates how to adapt NER for specialised use cases, making it more effective for business applications like regulatory compliance, customer insights and automated document processing.

Demo steps​

Custom entity configuration​

Set up the EntityRuler pipeline component and define patterns for domain-specific entities like government organisations, auto makers, financial institutions, tech companies and media outlets.

Apply custom NER to tweets​

Process tweets using the enhanced NER pipeline to extract mentions of custom entities, focusing on specific sectors and industries.

Analyse entity distributions​

Create visualisations to understand the frequency and distribution of custom entity types within the dataset.

Key takeaways​

  • Custom entity recognitionEntityRuler extends SpaCy’s NER by combining pattern-based matching with statistical models for specialised tasks.
  • Pattern definitionLabel-pattern pairs define entity types.
  • Case-insensitive matching improves recall.
  • Multiple patterns can map to the same entity.
  • Pipeline integrationEntityRuler runs before statistical NER, preserving original capabilities while adding domain-specific entities.
  • Entity analysisFrequency analysis, distribution insights and visualisation help track entity patterns across industries.