Open-source model small-e-czech-finetuned-ner-wikiann - Free support for Czech named entity recognition

Small E Czech Finetuned Ner Wikiann

Developed by richielo

This model is a fine-tuned version of Seznam/small-e-czech on the WikiANN dataset for Czech named entity recognition tasks.

Sequence Labeling

Transformers

#Czech NER #High F1 Score #WikiANN Fine-tuning

Downloads 10.07k

Release Time : 3/12/2022

Model Overview

This model is optimized for Czech named entity recognition (NER) tasks, fine-tuned on the WikiANN dataset, capable of identifying named entities in text.

Model Features

High Accuracy

Achieves an accuracy of 0.9557 on the Czech WikiANN test set.

Balanced Performance

Balanced precision (0.8713) and recall (0.8970), with an F1 score of 0.8840.

Specialized Optimization

Specifically fine-tuned for Czech named entity recognition tasks.

Model Capabilities

Czech Text Processing

Named Entity Recognition

Sequence Labeling

Use Cases

Text Analysis

News Entity Extraction

Identify entities such as person names, locations, and organizations in Czech news texts.

Helps news agencies automatically annotate key entities in news articles.

Document Processing

Process Czech documents to extract named entity information.

Can be used in document classification and information extraction systems.

🚀 small-e-czech-finetuned-ner-wikiann

This model is a fine - tuned version of [Seznam/small - e - czech](https://huggingface.co/Seznam/small - e - czech) on the wikiann dataset, achieving high performance in token classification tasks.

🚀 Quick Start

This model is a fine - tuned version of [Seznam/small - e - czech](https://huggingface.co/Seznam/small - e - czech) on the wikiann dataset. It achieves the following results on the evaluation set:

Loss: 0.2547
Precision: 0.8713
Recall: 0.8970
F1: 0.8840
Accuracy: 0.9557

📚 Documentation

Model Information

Property	Details
Model Type	small - e - czech - finetuned - ner - wikiann
Training Datasets	wikiann
Metrics	Precision, Recall, F1, Accuracy

Model Results

Task	Dataset	Metrics	Value
Token Classification	wikiann (cs)	Precision	0.8713322894683097
Token Classification	wikiann (cs)	Recall	0.8970423324922905
Token Classification	wikiann (cs)	F1	0.8840004144075699
Token Classification	wikiann (cs)	Accuracy	0.9557089381093997

🔧 Technical Details

Training Hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e - 05
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 20

Training Results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.2924	1.0	2500	0.2449	0.7686	0.8088	0.7882	0.9320
0.2042	2.0	5000	0.2137	0.8050	0.8398	0.8220	0.9400
0.1699	3.0	7500	0.1912	0.8236	0.8593	0.8411	0.9466
0.1419	4.0	10000	0.1931	0.8349	0.8671	0.8507	0.9488
0.1316	5.0	12500	0.1892	0.8470	0.8776	0.8620	0.9519
0.1042	6.0	15000	0.2058	0.8433	0.8811	0.8618	0.9508
0.0884	7.0	17500	0.2020	0.8602	0.8849	0.8724	0.9531
0.0902	8.0	20000	0.2118	0.8551	0.8837	0.8692	0.9528
0.0669	9.0	22500	0.2171	0.8634	0.8906	0.8768	0.9550
0.0529	10.0	25000	0.2228	0.8638	0.8912	0.8773	0.9545
0.0613	11.0	27500	0.2293	0.8626	0.8898	0.8760	0.9544
0.0549	12.0	30000	0.2276	0.8694	0.8958	0.8824	0.9554
0.0516	13.0	32500	0.2384	0.8717	0.8940	0.8827	0.9552
0.0412	14.0	35000	0.2443	0.8701	0.8931	0.8815	0.9554
0.0345	15.0	37500	0.2464	0.8723	0.8958	0.8839	0.9557
0.0412	16.0	40000	0.2477	0.8705	0.8948	0.8825	0.9552
0.0363	17.0	42500	0.2525	0.8742	0.8973	0.8856	0.9559
0.0341	18.0	45000	0.2529	0.8727	0.8962	0.8843	0.9561
0.0194	19.0	47500	0.2533	0.8699	0.8966	0.8830	0.9557
0.0247	20.0	50000	0.2547	0.8713	0.8970	0.8840	0.9557

Framework Versions

Transformers 4.17.0
Pytorch 1.10.0+cu111
Datasets 1.18.4
Tokenizers 0.11.6

📄 License

This model is released under the CC - BY - 4.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご