Natural language processing (NLP)

A technology that enables digital systems to understand and generate human language in written and verbal forms.

Natural language processing (NLP) has two related sub-fields: Natural Language Understanding (NLU) and Natural Language Generation (NLG).

Natural Language Processing stems from computer science, artificial intelligence, linguistics and data science. In practice, it uses machine learning and deep learning techniques. Those techniques allow NLPs to learn and improve their performance.

NLUs analyze the context of speech and text to understand the meaning of a sentence. That can help a system, such as a search engine, determine a user's intentions, based on what they type or say.

NLGs allow computers to generate words with meaning. Together, these technologies are used in chatbots and digital assistants. (za/fs)

Neural networks

In computer science and AI, neural networks are often considered similar to deep learning technologies — it just depends on which history of AI you follow.

As with deep learning, neural networks — or nets — are a type of AI model inspired by the human brain. They use large sets of data to learn skills, analyze information, and develop decision-making power.

AI neural networks are networks of thousands, sometimes millions, of interconnected nodes — points in a network through which data flows and connects with other data.

Neural networks often use hand-labelled examples from which they learn. For example, they are "fed" images of cars, labelled as cars, and learn what cars look like by detecting patterns, or car-like characteristics in those images.

A similar thing happens when you sign up to a new website and you are asked to select all the traffic lights in a set of images, or buses, bridges, crosswalks or motorcycles, before you're allowed to proceed. The software, reCAPTCHA, wants to make sure that you are human and stop bots creating masses of fake accounts. But the software has had to get more and more sophisticated, because every time we prove we are human, we help train the bots to be more human, too. Which is ironic.

While inspired by the brain, artificial neural networks are very different from biological nervous systems. They have different architecture, learning mechanisms, energy efficiency, and amounts of information they can integrate in tiny networks. According to many philosophers, artificial neural networks lack the capabilities of human reasoning, consciousness and emotions. (fs/za)

 

Coming soon:

No-code machine learning

Sources:

Machine Learning Glossary (Google) https://developers.google.com/machine-learning/glossary (accessed July 24, 2023)

NLP vs. NLU vs. NLG: the differences between three natural language processing concepts (IBM) https://www.ibm.com/blog/nlp-vs-nlu-vs-nlg-the-differences-between-three-natural-language-processing-concepts/ (accessed July 25, 2023)

Read the rest of DW's AI Lexicon:

A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z

We're keen to hear your feedback. Suggest an entry by sending us a comment. And let us know if you feel we have missed something, got it wrong, and tell us whether our AI Lexicon has helped you understand the technology better.

Written and edited by: Zulfikar Abbany (za), Fred Schwaller (fs)