prose
Pure Go natural language processing library for tokenization and tagging.
Tokenize, segment, tag, and extract named entities natively in Go without cgo or external services.
Prose is a pure Go natural language processing library designed for developers building text-processing applications. It provides foundational NLP tasks like tokenization, sentence segmentation, part-of-speech tagging, and named-entity extraction directly inside Go binaries. You should use it if you want zero external runtime dependencies and heavy control over your binary's memory footprint.
What it is
It parses text inputs to output token arrays, sentence boundaries, Universal Dependencies POS tags, and named entities with exact byte offsets. It supports 12 languages for tokenization, segmentation, and tagging, while keeping English-only packages for entity extraction and readability metrics. It also includes utility packages for text summarization readability scores and AP or Chicago style case conversion.
Who it's for
It suits Go developers who need embedded NLP pipelines without relying on Python bridges, cgo, or external network APIs. It does not suit teams working in non-Go ecosystems or those needing non-English named entity recognition out of the box.
Key features
Pure Go Tokenization
Processes words and punctuation with exact byte offsets in any language without cgo.
Universal Tagging Models
Averaged perceptron taggers emit Universal Dependencies tag sets across 12 languages.
Modular Import Design
Imports only required language packages and models to minimize binary sizes.
Readability and Casing
Calculates Flesch-Kincaid, Gunning fog, SMOG metrics, and handles title case conversion.
Up and running in 3 steps
Our honest verdict
What's great
- Operates entirely in pure Go without requiring cgo or external HTTP services.
- Retains exact byte offsets for every token, sentence, and entity relative to source strings.
- Modular package structure allows you to link only the specific language models you need.
- Provides context-aware processing functions for handling long texts safely.
Where it falls short
- Entity extraction and readability scoring are restricted to English inputs only.
- Requires writing Go code and managing compiled binaries rather than running as an out-of-the-box service.
- Language models must be imported explicitly, adding setup steps for multilingual pipelines.
Pricing & license
Free to use under the MIT license, with no hosted cloud option provided. The only cost is compiling and running it within your own Go applications.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
Does prose require cgo or external servers?
No, it is written entirely in pure Go and executes locally inside your application binary without external services.
Which languages are supported for part-of-speech tagging?
It supports twelve languages including English, Dutch, French, German, Spanish, and Italian using Universal Dependencies tags.
How can I reduce the binary size if I only need POS tags?
You can import the tag package directly instead of the root package to leave out entity recognition models.
