OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Ko Sentence Transformers

Ko Sentence Transformers is an open-source Python project that fine-tunes pretrained Korean language models on KorNLU datasets and publishes sentence-embedding models for use with sentence-transformers. It includes training and benchmarking scripts for models trained on KorNLI, KorSTS or both.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Ko Sentence Transformers — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a Python workflow needs sentence embeddings for Korean text and you want published models, training scripts or benchmark resources rather than a hosted inference endpoint.

Who it is for

  • Developers adding Korean semantic similarity or search to text workflows.
  • Teams building Korean text retrieval or sentence-ranking systems.
  • Practitioners who want to inspect or adapt KorNLU-based embedding training.

Documented uses

  • Encode Korean sentences as vectors for downstream NLP tasks.
  • Rank a corpus of sentences by similarity to a Korean query.
  • Build Korean text retrieval or clustering workflows using sentence embeddings.
  • Fine-tune Korean sentence-embedding models with the repository’s training approach.
Before you choose

Requirements & limits

The repository is public and lists the CC-BY-SA-4.0 license. Published models are available through Hugging Face for use with sentence-transformers.

  • The model variants differ by training data: KorNLI, KorSTS or both in multitask training.
  • The repository’s example uses sentence-transformers in Python and ranks results with cosine similarity.
  • Matryoshka variants are intended to work at multiple embedding dimensions; the README describes shorter vectors for candidate retrieval and full-dimension reranking.
  • The repository includes training_*.py and benchmark.py resources, while model inference is shown as a local Python workflow.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Encode and rank Korean sentences with a released model

Use a published model with sentence-transformers to rank corpus sentences by similarity to a Korean query.

Before you start

  • Python
  • sentence-transformers
  • A published model such as jhgan/ko-sroberta-multitask
  • A collection of Korean sentences and a Korean query

Steps

  1. Load the model with SentenceTransformer("jhgan/ko-sroberta-multitask").

  2. Encode the corpus sentences as tensor embeddings.

  3. Encode a query sentence and calculate cosine similarity against the corpus embeddings.

  4. Select and display the corpus sentences with the highest similarity scores.

Expected outcome

A similarity-ranked list of corpus sentences for each query.

Guide limitations

  • The repository example demonstrates sentence-level similarity ranking; it does not document a hosted inference endpoint.
  • Similarity scores rank the example corpus and are not presented as guarantees of retrieval quality for other data.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
jhgan00/ko-sentence-transformers · Official homepage title · retained in its source language
How we review sourcesSuggest a correction