Definition
Natural Language Processing (NLP) covers tasks like translation, summarization, sentiment analysis, named entity recognition, and question answering — anything involving text or speech as input or output.
Modern NLP is dominated by deep learning and LLMs, but classical methods (regex, TF-IDF, HMMs) still appear in pipelines.
In simple terms
NLP is teaching computers to read, listen, and write in human languages — not just match keywords, but handle grammar, meaning, and context.
Where you see it
- Google Translate and Somali–English MT systems.
- Spam filters and sentiment analysis on social media.
- Goobo Labs builds Somali NLP datasets, models, and benchmarks.
How it works
1.Acquire text
Collect or crawl corpora in the target language.
2.Preprocess
Tokenize, normalize, and optionally tag language or dialect.
3.Model
Train or prompt classifiers, seq2seq models, or LLMs for the task.
4.Evaluate
Measure accuracy, BLEU, F1, or human preference on test sets.
Why it matters
- NLP is how AI interfaces with human communication — the core of chatbots, search, and translation.
Often confused
LLMs replaced all of NLP.
LLMs are one tool; retrieval, rules, and smaller models still matter for cost, latency, and control.