It is a RoBERTa model trained for POS-tagging and dependency-parsing (using goeswith for subwords), optimized for deployment in JavaScript/TypeScript environments (Node.js, Deno, or directly in the browser) without requiring Python.

How to Use with TypeScript

You can easily run this model in your modern TypeScript stack using Transformers.js.

typescript

import { pipeline } from "@huggingface/transformers";

// Initialize the token classification pipeline targeting this specific ONNX repository
const modelId = "ghotriw/roberta-base-english-ud-goeswith-onnx";
const tagger = await pipeline("token-classification", modelId);

// Use aggregation_strategy: 'none' to get raw labels for each subword
const rawTokens = await tagger("I saw a horse", { aggregation_strategy: 'none' });

for (const token of rawTokens) {
  const [pos, features, deprel] = token.entity.split('|');
  console.log(`Word: ${token.word}, POS: ${pos}, Dep: ${deprel}`);
}