Making computers work with human language.
WHAT IT COVERS
Understanding, generating and transforming text and speech.
WHAT IT INCLUDES
Translation Summarisation Sentiment analysis Named entity recognition, identifying people, places and organisations Question answering Speech recognition and synthesis Classification
WHERE YOU ENCOUNTER IT
Spam filtering Search engines Voice assistants Autocomplete Customer service routing Translation tools
HOW IT CHANGED
Earlier approaches used hand-written rules and were brittle.
Statistical methods improved matters. Deep learning transformed them.
Modern language models handle most of these tasks without task-specific training.
WHAT IS STILL HARD
Sarcasm and irony Context requiring world knowledge Low-resource languages with little training data Domain-specific terminology
FOR NIGERIAN LANGUAGES
Support varies considerably. Widely spoken languages have better support than others.
Test before relying on translation for anything important.
THE PRACTICAL POINT
Language tasks are where current AI is strongest.
