Practical overview
When it fits
Use it when a project needs Korean text emotion labels or a dataset and model baseline for Korean multi-label emotion-classification work.
Who it is for
- Developers adding Korean emotion classification to a text-processing workflow.
- Researchers or practitioners comparing Korean emotion-classification models using the provided dataset and evaluation code.
Documented uses
- Assign one or more of 28 emotion labels and scores to Korean text.
- Train or evaluate KoELECTRA-based models with the supplied configurations.
- Use the translated dataset as a basis for emotion-classification experiments.
Requirements & limits
- The classifier uses sigmoid-based multi-label output with a documented threshold of 0.3.
- Four documented model variants cover KoELECTRA-v1 and v3 in small and base sizes.
- The README lists an older, specific Python dependency stack, which matters when reproducing its examples.
- The repository is licensed under Apache-2.0.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Run the documented Korean emotion inference pipeline
Load a published GoEmotions Korean model and classify a list of Korean text strings.
Before you start
- Python environment with the repository’s listed dependencies: torch==1.7.1, transformers==3.5.1, googletrans==2.4.1 and attrdict==2.0.1.
- The repository’s custom MultiLabelPipeline, model and tokenizer code.
Steps
Import MultiLabelPipeline, ElectraTokenizer and ElectraForMultiLabelClassification as shown in the README.
Load the tokenizer and model from a documented model identifier, such as monologg/koelectra-base-v3-goemotions.
Create MultiLabelPipeline with the model, tokenizer and threshold 0.3.
Pass a list of Korean text strings to the pipeline to obtain labels and scores.
Expected outcome
A list of predicted emotion labels and associated scores for each input text; multiple labels may be returned.
Guide limitations
- The repository’s translated Reddit data is described by its maintainers as poor in translation quality.
- The README’s examples and dependency versions are specific to its documented Transformers-era implementation.
Train or evaluate a model with the repository configuration
Use the repository’s supplied configuration files to start its documented training and evaluation workflow.
Before you start
- Python 3 and the repository’s listed requirements.
- The repository’s data, code and configuration files.
Steps
Install the listed requirements using the repository’s requirements file.
Choose a supplied configuration, such as koelectra-base.json or koelectra-small.json.
Run run_goemotions.py with the chosen configuration file.
Expected outcome
The documented training and evaluation workflow runs using the selected KoELECTRA configuration.
Guide limitations
- The repository warns that translated data quality is poor.
- The documented commands specify the workflow but do not provide a guarantee of results in a particular environment.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 17 October 2026
- Original website title
- monologg/GoEmotions-Korean · Official homepage title · retained in its source language