text corpora
Text corpora (singular corpus) are collections of machine-readable text that serve as foundational data resources in natural language processing (NLP). They range from small, carefully curated collections assembled by linguists to web-scale crawls of hundreds of billions of pages that are filtered and deduplicated rather than hand-picked. They may be raw (unannotated) or annotated with linguistic structure, such as part-of-speech tags, lemmas, syntactic trees, and semantic roles.
Common corpus types include:
- Monolingual corpora containing text in a single language
- Multilingual corpora containing text in multiple languages
- Parallel corpora with aligned texts in two or more languages, which is useful for translation
- Comparable corpora with texts in multiple languages covering comparable topics but not direct translations
- Balanced corpora designed to reflect diverse genres, registers, or subdomains in proportion
- Domain-specific or specialized corpora focused on a particular field, genre, or source, such as medical texts, legal documents, or tweets
Text corpora are used to study real-world language use, inform linguistic analyses, train and evaluate computational models, and benchmark tasks, such as tagging, parsing, machine translation, classification, retrieval, and generation. In modern AI work, the largest text corpora are the web-scale pretraining datasets that supply the trillions of tokens used to pretrain large language models (LLMs). These datasets come from crawls like Common Crawl and are released as filtered, deduplicated collections such as C4, The Pile, and FineWeb.
Related Resources
Tutorial
Natural Language Processing With Python's NLTK Package
In this beginner-friendly tutorial, you'll take your first steps with Natural Language Processing (NLP) and Python's Natural Language Toolkit (NLTK). You'll learn how to process unstructured data in order to be able to analyze it and draw conclusions from it.
For additional information on related topics, take a look at the following resources:
By Leodanis Pozo Ramos • Updated Aug. 2, 2026
