Open-source ko_fin_ner_roberta_small_model - Accurately identify specific entities in Korean financial texts

Ko Fin Ner Roberta Small Model

Developed by Hyeonseo

A Korean financial domain named entity recognition model fine-tuned from klue/roberta-small, identifying specific entities in financial texts

Sequence Labeling

Transformers

Korean#Korean Financial Entity Recognition #High Recall NER #Roberta-small Fine-tuning

Downloads 16

Release Time : 7/16/2023

Model Overview

This model is specifically optimized for Korean financial texts, capable of recognizing named entities in the financial domain such as company names and financial products.

Model Features

Financial Domain Optimization

Fine-tuned specifically for Korean financial texts, optimizing entity recognition capabilities in the financial domain

Efficient Performance

Based on the small RoBERTa model, maintaining high performance while reducing computational resource requirements

Multi-metric Evaluation

Provides multi-dimensional performance evaluations including precision, recall, F1-score, and accuracy

Model Capabilities

Korean Text Processing

Financial Entity Recognition

Named Entity Labeling

Use Cases

Financial Analysis

Financial News Entity Extraction

Identifying companies, products, and market entities from financial news

F1-score reached 0.8050

Securities Report Analysis

Analyzing key entity information in securities research reports

🚀 ko_fin_ner_roberta_small_model

This model is a fine - tuned version of klue/roberta-small. It offers a high - performance solution for specific tasks in the financial domain, leveraging the pre - trained features of the base model and achieving excellent results on the evaluation set.

📚 Documentation

Model Information

Property	Details
Language	Korean
Tags	Generated from trainer
Metrics	Precision, Recall, F1, Accuracy
Base Model	klue/roberta - small
Model Name	ko_fin_ner_roberta_small_model

Model Performance

This model is a fine - tuned version of [klue/roberta - small](https://huggingface.co/klue/roberta - small) on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.2873
Precision: 0.7436
Recall: 0.8774
F1: 0.8050
Accuracy: 0.9374

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e - 05
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9, 0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	25	1.0272	0.1215	0.1662	0.1404	0.7237
No log	2.0	50	0.7136	0.2360	0.4033	0.2978	0.7695
No log	3.0	75	0.5289	0.3422	0.5586	0.4244	0.8285
No log	4.0	100	0.4404	0.4184	0.6076	0.4956	0.8730
No log	5.0	125	0.3768	0.4124	0.6540	0.5058	0.8866
No log	6.0	150	0.3484	0.4758	0.6975	0.5657	0.8953
No log	7.0	175	0.3236	0.5477	0.7357	0.6279	0.9039
No log	8.0	200	0.3097	0.5702	0.7520	0.6486	0.9015
No log	9.0	225	0.3168	0.6167	0.7629	0.6821	0.9096
No log	10.0	250	0.2950	0.6176	0.8011	0.6975	0.9145
No log	11.0	275	0.2806	0.6674	0.8147	0.7337	0.9195
No log	12.0	300	0.2749	0.6853	0.8365	0.7534	0.9266
No log	13.0	325	0.2743	0.7002	0.8338	0.7612	0.9292
No log	14.0	350	0.2862	0.6774	0.8011	0.7341	0.9238
No log	15.0	375	0.2703	0.6879	0.8529	0.7616	0.9276
No log	16.0	400	0.2752	0.7036	0.8474	0.7689	0.9293
No log	17.0	425	0.2721	0.6998	0.8447	0.7654	0.9305
No log	18.0	450	0.2831	0.6979	0.8311	0.7587	0.9299
No log	19.0	475	0.2857	0.7252	0.8556	0.7850	0.9319
0.2786	20.0	500	0.2792	0.7260	0.8665	0.7901	0.9319
0.2786	21.0	525	0.2604	0.7355	0.8638	0.7945	0.9349
0.2786	22.0	550	0.2603	0.7092	0.8638	0.7789	0.9359
0.2786	23.0	575	0.3026	0.7227	0.8665	0.7881	0.9342
0.2786	24.0	600	0.2800	0.7431	0.8747	0.8035	0.9375
0.2786	25.0	625	0.2838	0.7283	0.8692	0.7925	0.9361
0.2786	26.0	650	0.2813	0.7339	0.8719	0.7970	0.9371
0.2786	27.0	675	0.2881	0.7407	0.8719	0.8010	0.9358
0.2786	28.0	700	0.2894	0.7379	0.8747	0.8005	0.9362
0.2786	29.0	725	0.2889	0.7483	0.8747	0.8065	0.9368
0.2786	30.0	750	0.2873	0.7436	0.8774	0.8050	0.9374

Framework versions

Transformers 4.28.0
Pytorch 2.0.1+cu118
Datasets 2.13.1
Tokenizers 0.13.3

Example Inputs

example01: "나스닥투자증권에서 시작된 발동성 가치 상태 효과는 투자자들에게 좋은 기회를 제공합니다."
example02: "TM머니가 베를린증권거래소에서 미국 보험 유가를 거래하고 있습니다."

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