RecursiveAI.OneNlp.Models.Multilingual 1.0.0

Prefix Reserved
dotnet add package RecursiveAI.OneNlp.Models.Multilingual --version 1.0.0
                    
NuGet\Install-Package RecursiveAI.OneNlp.Models.Multilingual -Version 1.0.0
                    
This command is intended to be used within the Package Manager Console in Visual Studio, as it uses the NuGet module's version of Install-Package.
<PackageReference Include="RecursiveAI.OneNlp.Models.Multilingual" Version="1.0.0" />
                    
For projects that support PackageReference, copy this XML node into the project file to reference the package.
<PackageVersion Include="RecursiveAI.OneNlp.Models.Multilingual" Version="1.0.0" />
                    
Directory.Packages.props
<PackageReference Include="RecursiveAI.OneNlp.Models.Multilingual" />
                    
Project file
For projects that support Central Package Management (CPM), copy this XML node into the solution Directory.Packages.props file to version the package.
paket add RecursiveAI.OneNlp.Models.Multilingual --version 1.0.0
                    
#r "nuget: RecursiveAI.OneNlp.Models.Multilingual, 1.0.0"
                    
#r directive can be used in F# Interactive and Polyglot Notebooks. Copy this into the interactive tool or source code of the script to reference the package.
#:package RecursiveAI.OneNlp.Models.Multilingual@1.0.0
                    
#:package directive can be used in C# file-based apps starting in .NET 10 preview 4. Copy this into a .cs file before any lines of code to reference the package.
#addin nuget:?package=RecursiveAI.OneNlp.Models.Multilingual&version=1.0.0
                    
Install as a Cake Addin
#tool nuget:?package=RecursiveAI.OneNlp.Models.Multilingual&version=1.0.0
                    
Install as a Cake Tool

RecursiveAI.OneNLP.Models.Multilingual

spaCy-class NLP for .NET. The distilled accurate-tier model for RecursiveAI.OneNlp: one six-layer multilingual encoder (a student of mmBERT-base, fine-tuned) shared by heads for English, German, French, Spanish, Portuguese, Italian, Dutch, Turkish and Chinese (tags, morphology, lemmas, dependency parse, entities), and an open-type entity recognizer whose entity types you name at run time. Add this package and the Nlp verbs use it for those languages whenever no fast-tier package for the language is referenced; a fast-tier package, when present, is used instead. If RecursiveAI.OneNlp.Models.Multilingual.Large is referenced too, the accurate tier uses that. No downloads, no native code, NativeAOT and trim clean. Unaffiliated with other projects named OneNLP.

Measured. The distilled tier is an information tier, not the base-size accurate-tier quality gate. On English, the fine-tuned student scores UPOS 95.88 [95.55, 96.19] on UD English-EWT r2.18 test (gold tokens); the frozen-encoder entity path scores micro F1 75.48 [72.66, 78.03] on UNER English-EWT test. The published UDPipe 2 UPOS figure is 97.61 (UD 2.17) and XLM-R large entity F1 is 85.8; these are different-release/model references, not paired comparisons. The full language rows and protocol notes are in docs/results-table.md and the accurate-tier recipe results. Data lineage: UD r2.18 and entity corpora (CC BY-SA and CC BY); mmBERT-base (MIT) distilled on FineWeb-2 and Wikipedia, so the weights carry share-alike lineage.

Open-type NER on CrossNER zero-shot test scores F1 41.27 (politics), 36.50 (science), 26.01 (music), 30.38 (literature) and 21.10 (AI), 19–35 points below the published GLiNER-S numbers. Protocol equivalence is unverified and no zero-shot quality pass is claimed. See docs/open-ner.md for the audit.

Weights total about 505 MiB across this package and its three parts (...Multilingual.Encoder, .FrozenEncoder, .Segmenters); reference only this package. The frozen encoder is also usable on its own, as RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder (FrozenStudentEncoder.Create()).

using RecursiveAI.OneNlp;

Doc doc = Nlp.Analyze("Angela Merkel besuchte im Mai die Stadt Hamburg.");
foreach (Token t in doc)
{
    Console.WriteLine($"{t} {t.Pos} {t.Lemma} {t.Dep}");
}

// Entity types named at run time:
foreach (Span e in Nlp.Entities("Acme Corp. signed the lease with Beta LLC.", "organization"))
{
    Console.WriteLine($"{e.Label} {e}");
}

// Unusual type names score lower; lower the threshold (default 0.5) and check what comes back:
IReadOnlyList<Span> parties = Nlp.Entities("Acme Corp. signed the lease with Beta LLC.", ["contract party"], threshold: 0.1f);

Chinese words and Turkish multiword tokens come from small trained tokenizers in the package. The weights are trained on Universal Dependencies treebanks and entity corpora (several CC BY-SA), credited in THIRD-PARTY-NOTICES.md; PROVENANCE.md lists every dataset, recipe and seed. The weights are licensed MIT OR Apache-2.0.

Copyright (c) Recursive AI, LLC.

Product Compatible and additional computed target framework versions.
.NET net10.0 is compatible.  net10.0-android was computed.  net10.0-browser was computed.  net10.0-ios was computed.  net10.0-maccatalyst was computed.  net10.0-macos was computed.  net10.0-tvos was computed.  net10.0-windows was computed. 
Compatible target framework(s)
Included target framework(s) (in package)
Learn more about Target Frameworks and .NET Standard.

NuGet packages

This package is not used by any NuGet packages.

GitHub repositories

This package is not used by any popular GitHub repositories.

Version Downloads Last Updated
1.0.0 114 11/8/2026

See CHANGELOG.md in the repository.