https://github.com/stefan-it/georgian-ner
Resources about Named Entity Recognition for Georgian
https://github.com/stefan-it/georgian-ner
flair georgian named-entity-recognition
Last synced: about 2 months ago
JSON representation
Resources about Named Entity Recognition for Georgian
- Host: GitHub
- URL: https://github.com/stefan-it/georgian-ner
- Owner: stefan-it
- Created: 2023-11-15T21:09:56.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2023-11-17T01:12:47.000Z (over 1 year ago)
- Last Synced: 2025-02-08T07:13:23.525Z (3 months ago)
- Topics: flair, georgian, named-entity-recognition
- Language: Jupyter Notebook
- Homepage:
- Size: 91.8 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🇬🇪 Georgian NER
გამარჯობა! This repository contains my resources about Named Entity Recognition for Georgian.
# English-Georgian NER Model with Flair
We fine-tune a NER Model with Flair on English and Georgian training splits of WikiANN dataset
([Rahimi et al.](https://www.aclweb.org/anthology/P19-1015) splits).
Based on [this repository](https://github.com/stefan-it/autotrain-flair-mobie) the fine-tuning is done with the
awesome [Flair](https://github.com/flairNLP/flair) library, incl. support for Hugging Face's [AutoTrain](https://github.com/huggingface/autotrain-advanced).
We use a basic hyper-parameter search with the following configuration:
| Parameter | Value |
|---------------|-------------------|
| Learning Rate | `5e-06` |
| Batch Size | `4` |
| Epoch | `10` |
| Seeds | `[1, 2, 3, 4, 5]` |
We use [XLM-R Large](https://huggingface.co/xlm-roberta-large) as base model.
The following environment variables needs to be set when using AutoTrain:
| Environment Variable | Description |
|----------------------|---------------------------------------------------------------------------------------------------|
| `HF_TOKEN` | Hugging Face User Access Token, which can be found [here](https://huggingface.co/settings/tokens) |
| `HUB_ORG_NAME` | Username or organization under models will be uploaded to |
The fine-tuning can then be started by running the `script.py` script.
## Model Card Creation
The [`ModelCardCreation.ipynb`](ModelCardCreation.ipynb) notebook shows how to automatically generate model cards for
all uploaded models. This includes also a results overview table with linked models.
## Fine-tuned Models
All fine-tuned models are released on the Hugging Face Hub, incl. a nice inference widget:
The fine-tuned models can be found [here](https://huggingface.co/models?search=autotrain-flair-georgian). Additionally,
they can be found in [this collection](https://huggingface.co/collections/stefan-it/georgian-ner-models-6556bd33dd1c096392074791).
# Changelog
* 17.11.2023: Add model card creation and fine-tuned models sections. Mention fine-tuned models on Hub.
* 15.11.2023: Initial version of this repository.