Named Entity Recognition (NER), also known as entity extraction or chunking, is a foundational Natural Language Processing (NLP) technique aimed at the extraction and classification of key information (named entities) from text data. It is classified as a supervised ML, prediction-based approach, meaning the algorithm is trained on labeled examples to make predictions on unseen data. The process takes unstructured text (like a sentence, paragraph, or entire document) and identifies words or phrases, labeling them into predefined categories such as person, organization, location, event, and date.
Entity extraction is important because it adds structure and semantic information to previously unstructured text, which is a vital preprocessing step for numerous downstream NLP tasks, including those performed by Large Language Models. The output typically includes the entity name and its type, along with rich metrics such as salience (importance), sentiment score/magnitude, number of mentions, and metadata (like Knowledge Graph identifiers). This comprehensive data is critical for advanced analyses like semantic keyword research and internal linking audits.