Kobert: Korean BERT for sentiment analysis

Understand the project

Practical overview

Use KoBERT when a project needs a pretrained Korean BERT component and its documented framework examples match the development environment.

Developers building Korean-language NLP workflows.
Machine-learning practitioners fine-tuning a model for Korean text classification or named-entity recognition.

Tokenize Korean text with the provided SentencePiece vocabulary.
Generate sequence encodings and pooled model representations.
Use the model as a base for task-specific Korean sentiment analysis or entity recognition.

Before you choose

The public GitHub repository specifies the Apache-2.0 license.

The repository documents PyTorch, ONNX Runtime and MXNet-Gluon usage.
The tokenizer vocabulary contains 8,002 entries and is based on SentencePiece.
The model defaults to evaluation mode; the README says to switch to training mode for fine-tuning.
The README describes a 12-layer architecture and maximum sequence length of 512.

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

Add KoBERT as a Python package, or prepare a source checkout with the listed requirements.

Install the package with the command below.
Documented command
```
pip install git+https://git@github.com/SKTBrain/KoBERT.git@master
```
To modify source code instead, clone https://github.com/SKTBrain/KoBERT.git and enter the KoBERT directory.
Install the checkout's dependencies with the command below.
Documented command
```
pip install -r requirements.txt
```

KoBERT is installed as a Python package, or its source is checked out with the repository requirements installed.

These instructions cover installation, not a complete application or deployment.
The repository directs users to requirements.txt for dependencies.

Evidence and transparency

Source record details

Source availability: Active
Sources checked: 15 October 2026
Original website title: SKTBrain/KoBERT · Official homepage title · retained in its source language