Letter N

Finding and labelling names such as people, places, companies or dates in text.

It is like using different highlighter colours for different kinds of name. The system first finds the words, then decides what type of thing each name represents.

#

Using a computer system to produce human-readable language.

NLG covers simple template-based reports as well as flexible text from language models. It is the writing side of language technology, while NLP is the wider field of working with language.

#

The field of making computers work with human language.

NLP covers reading, writing, translation, speech, search, classification and more. Language models are part of NLP, but the field existed long before modern chatbots.

#

A method that learns a three-dimensional scene from photographs taken at different viewpoints.

Imagine walking around an object and taking many pictures. A NeRF learns how colour and light should appear at points in that space, allowing it to create convincing views from camera positions that were never photographed directly.

#

A learning system made of connected mathematical units arranged in layers.

Neural networks are loosely inspired by brains but are not digital brains. During training, their connections are adjusted so useful patterns produce better outputs.

#

Capable of producing different results from the same input.

Many generative models choose among several plausible next tokens, so repeated requests can vary. Settings and system design can reduce variation but may not remove it completely.

#