NLP

Natural Language Processing (NLP) is the broader field covering all computational work with human language, including both understanding and generation. Natural Language Understanding (NLU) is a subset of NLP focused specifically on interpreting the meaning, intent, and context of input text — essentially the comprehension side. Natural Language Generation (NLG) is the complementary subset that covers producing text from structured data or internal representations.

Before large language models, NLP tasks required separate specialised models — one for sentiment analysis, another for translation, another for named entity recognition. LLMs unified NLP by demonstrating that a single large pre-trained model can perform all these tasks through prompting or minimal fine-tuning. This paradigm shift, beginning with BERT (2018) and accelerating with GPT-3 (2020), reduced the need for task-specific feature engineering and custom architectures.

NLP works with any natural language, but model quality varies significantly by language. English is the most well-resourced language in NLP — the majority of training data and benchmarks are in English. Multilingual models like mBERT and NLLB-200 extend NLP capabilities to 100+ languages, but low-resource languages (those with limited digital text) still see substantially lower performance than English. Languages with non-Latin scripts or complex morphology (e.g. Arabic, Chinese, Finnish) present additional modelling challenges.