Hugging Face's logo

Leo97
/
KoELECTRA-small-v3-modu-ner

Token Classification
Transformers
PyTorch
TensorBoard
Safetensors
Korean
electra
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
2
Leo97 commited on
Commit
e3e398a
•
1 Parent(s): 692d858

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +34 -79
README.md CHANGED
@@ -9,77 +9,32 @@ metrics:
9
  model-index:
10
  - name: KoELECTRA-small-v3-modu-ner
11
  results: []
12
- language:
13
- - ko
14
- pipeline_tag: token-classification
15
- widget:
16
- - text: "서울역으로 안내해줘."
17
- example_title: "Example 1"
18
- - text: "에어컨 온도 3도 올려줘."
19
- example_title: "Example 2"
20
- - text: "아이유 노래 검색해줘."
21
- example_title: "Example 3"
22
  ---
23
 
 
 
 
24
  # KoELECTRA-small-v3-modu-ner
25
 
26
  This model is a fine-tuned version of [monologg/koelectra-small-v3-discriminator](https://huggingface.co/monologg/koelectra-small-v3-discriminator) on an unknown dataset.
27
  It achieves the following results on the evaluation set:
28
- - Loss: 0.1443
29
- - Precision: 0.8176
30
- - Recall: 0.8401
31
- - F1: 0.8287
32
- - Accuracy: 0.9615
33
 
34
  ## Model description
35
 
36
- 태깅 시스템 : BIO 시스템
37
- - B-(begin) : 개체명이 시작할 때
38
- - I-(inside) : 토큰이 개체명 중간에 있을 때
39
- - O(outside) : 토큰이 개체명이 아닐 경우
40
-
41
- 한국정보통신기술협회(TTA) 대분류 기준을 따르는 15 가지의 태그셋
42
-
43
- | 분류 | 표기 | 정의 |
44
- |:------------:|:---:|:-----------|
45
- | ARTIFACTS | AF | 사람에 의해 창조된 인공물로 문화재, 건물, 악기, 도로, 무기, 운송수단, 작품명, 공산품명이 모두 이에 해당 |
46
- | ANIMAL | AM | 사람을 제외한 짐승 |
47
- | CIVILIZATION | CV | 문명/문화 |
48
- | DATE | DT | 기간 및 계절, 시기/시대 |
49
- | EVENT | EV | 특정 사건/사고/행사 명칭 |
50
- | STUDY_FIELD | FD | 학문 분야, 학파 및 유파 |
51
- | LOCATION | LC | 지역/장소와 지형/지리 명칭 등을 모두 포함 |
52
- | MATERIAL | MT | 원소 및 금속, 암석/보석, 화학물질 |
53
- | ORGANIZATION | OG | 기관 및 단체 명칭 |
54
- | PERSON | PS | 인명 및 인물의 별칭 (유사 인물 명칭 포함) |
55
- | PLANT | PT | 꽃/나무, 육지식물, 해초류, 버섯류, 이끼류 |
56
- | QUANTITY | QT | 수량/분량, 순서/순차, 수사로 이루어진 표현 |
57
- | TIME | TI | 시계상으로 나타나는 시/시각, 시간 범위 |
58
- | TERM | TM | 타 개체명에서 정의된 세부 개체명 이외의 개체명 |
59
- | THEORY | TR | 특정 이론, 법칙 원리 등 |
60
 
61
  ## Intended uses & limitations
62
 
63
- ### How to use
64
- You can use this model with Transformers *pipeline* for NER.
65
- ```python
66
- from transformers import AutoTokenizer, AutoModelForTokenClassification
67
- from transformers import pipeline
68
-
69
- tokenizer = AutoTokenizer.from_pretrained("Leo97/KoELECTRA-small-v3-modu-ner")
70
- model = AutoModelForTokenClassification.from_pretrained("Leo97/KoELECTRA-small-v3-modu-ner")
71
- ner = pipeline("ner", model=model, tokenizer=tokenizer)
72
-
73
- example = "서울역으로 안내해줘."
74
- ner_results = ner(example)
75
- print(ner_results)
76
- ```
77
 
78
  ## Training and evaluation data
79
 
80
- 개체명 인식(NER) 모델 학습 데이터 셋
81
- - 문화체육관광부 > 국립국어원 > 모두의 말뭉치 > 개체명 분석 말뭉치 2021
82
- - https://corpus.korean.go.kr/request/reausetMain.do
83
 
84
  ## Training procedure
85
 
@@ -92,34 +47,34 @@ The following hyperparameters were used during training:
92
  - seed: 42
93
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
94
  - lr_scheduler_type: linear
95
- - lr_scheduler_warmup_steps: 3787
96
- - num_epochs: 18 (= 10 + 3 + 5)
97
  - mixed_precision_training: Native AMP
98
 
99
  ### Training results
100
 
101
  | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
102
  |:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:|
103
- | No log | 1.0 | 3788 | 0.3021 | 0.6356 | 0.6380 | 0.6368 | 0.9223 |
104
- | No log | 2.0 | 7576 | 0.1905 | 0.7397 | 0.7441 | 0.7419 | 0.9431 |
105
- | No log | 3.0 | 11364 | 0.1612 | 0.7611 | 0.7897 | 0.7751 | 0.9505 |
106
- | No log | 4.0 | 15152 | 0.1494 | 0.7855 | 0.7998 | 0.7926 | 0.9544 |
107
- | No log | 5.0 | 18940 | 0.1427 | 0.7833 | 0.8194 | 0.8009 | 0.9559 |
108
- | No log | 6.0 | 22728 | 0.1398 | 0.7912 | 0.8223 | 0.8064 | 0.9572 |
109
- | No log | 7.0 | 26516 | 0.1361 | 0.8035 | 0.8240 | 0.8136 | 0.9587 |
110
- | No log | 8.0 | 30304 | 0.1360 | 0.8047 | 0.8280 | 0.8162 | 0.9592 |
111
- | No log | 9.0 | 34092 | 0.1346 | 0.8058 | 0.8299 | 0.8177 | 0.9596 |
112
- | 0.2256 | 10.0 | 37880 | 0.1350 | 0.8068 | 0.8308 | 0.8186 | 0.9598 |
113
- | 3회 훈련 추가 ||||||||
114
- | No log | 1.0 | 3788 | 0.1367 | 0.8089 | 0.8240 | 0.8164 | 0.9595 |
115
- | No log | 2.0 | 7576 | 0.1345 | 0.8130 | 0.8331 | 0.8229 | 0.9604 |
116
- | 0.0953 | 3.0 | 11364 | 0.1370 | 0.8146 | 0.8349 | 0.8246 | 0.9609 |
117
- | 5회 훈련 추가 ||||||||
118
- | No log | 1.0 | 3788 | 0.1511 | 0.8095 | 0.8257 | 0.8176 | 0.9594 |
119
- | No log | 2.0 | 7576 | 0.1461 | 0.8121 | 0.8339 | 0.8228 | 0.9600 |
120
- | No log | 3.0 | 11364 | 0.1417 | 0.8139 | 0.8372 | 0.8254 | 0.9607 |
121
- | No log | 4.0 | 15152 | 0.1418 | 0.8238 | 0.8346 | 0.8292 | 0.9617 |
122
- | 0.0748 | 5.0 | 18940 | 0.1443 | 0.8176 | 0.8401 | 0.8287 | 0.9615 |
123
 
124
 
125
  ### Framework versions
@@ -127,4 +82,4 @@ The following hyperparameters were used during training:
127
  - Transformers 4.27.4
128
  - Pytorch 2.0.0+cu118
129
  - Datasets 2.11.0
130
- - Tokenizers 0.13.2
 
9
  model-index:
10
  - name: KoELECTRA-small-v3-modu-ner
11
  results: []
 
 
 
 
 
 
 
 
 
 
12
  ---
13
 
14
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
+ should probably proofread and complete it, then remove this comment. -->
16
+
17
  # KoELECTRA-small-v3-modu-ner
18
 
19
  This model is a fine-tuned version of [monologg/koelectra-small-v3-discriminator](https://huggingface.co/monologg/koelectra-small-v3-discriminator) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Loss: 0.1431
22
+ - Precision: 0.8232
23
+ - Recall: 0.8449
24
+ - F1: 0.8339
25
+ - Accuracy: 0.9628
26
 
27
  ## Model description
28
 
29
+ More information needed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
30
 
31
  ## Intended uses & limitations
32
 
33
+ More information needed
 
 
 
 
 
 
 
 
 
 
 
 
 
34
 
35
  ## Training and evaluation data
36
 
37
+ More information needed
 
 
38
 
39
  ## Training procedure
40
 
 
47
  - seed: 42
48
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
49
  - lr_scheduler_type: linear
50
+ - lr_scheduler_warmup_steps: 15151
51
+ - num_epochs: 20
52
  - mixed_precision_training: Native AMP
53
 
54
  ### Training results
55
 
56
  | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
57
  |:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:|
58
+ | No log | 1.0 | 3788 | 0.3978 | 0.5986 | 0.5471 | 0.5717 | 0.9087 |
59
+ | No log | 2.0 | 7576 | 0.2319 | 0.6986 | 0.6953 | 0.6969 | 0.9345 |
60
+ | No log | 3.0 | 11364 | 0.1838 | 0.7363 | 0.7612 | 0.7486 | 0.9444 |
61
+ | No log | 4.0 | 15152 | 0.1610 | 0.7762 | 0.7745 | 0.7754 | 0.9509 |
62
+ | No log | 5.0 | 18940 | 0.1475 | 0.7862 | 0.8011 | 0.7936 | 0.9545 |
63
+ | No log | 6.0 | 22728 | 0.1417 | 0.7857 | 0.8181 | 0.8016 | 0.9563 |
64
+ | No log | 7.0 | 26516 | 0.1366 | 0.8022 | 0.8196 | 0.8108 | 0.9584 |
65
+ | No log | 8.0 | 30304 | 0.1346 | 0.8093 | 0.8236 | 0.8164 | 0.9596 |
66
+ | No log | 9.0 | 34092 | 0.1328 | 0.8085 | 0.8299 | 0.8190 | 0.9602 |
67
+ | No log | 10.0 | 37880 | 0.1332 | 0.8110 | 0.8368 | 0.8237 | 0.9608 |
68
+ | No log | 11.0 | 41668 | 0.1323 | 0.8157 | 0.8347 | 0.8251 | 0.9612 |
69
+ | No log | 12.0 | 45456 | 0.1353 | 0.8118 | 0.8402 | 0.8258 | 0.9611 |
70
+ | No log | 13.0 | 49244 | 0.1370 | 0.8152 | 0.8416 | 0.8282 | 0.9616 |
71
+ | No log | 14.0 | 53032 | 0.1368 | 0.8164 | 0.8415 | 0.8287 | 0.9616 |
72
+ | No log | 15.0 | 56820 | 0.1378 | 0.8187 | 0.8438 | 0.8310 | 0.9621 |
73
+ | No log | 16.0 | 60608 | 0.1389 | 0.8217 | 0.8438 | 0.8326 | 0.9626 |
74
+ | No log | 17.0 | 64396 | 0.1380 | 0.8266 | 0.8426 | 0.8345 | 0.9631 |
75
+ | No log | 18.0 | 68184 | 0.1428 | 0.8216 | 0.8445 | 0.8329 | 0.9625 |
76
+ | No log | 19.0 | 71972 | 0.1431 | 0.8232 | 0.8455 | 0.8342 | 0.9628 |
77
+ | 0.1712 | 20.0 | 75760 | 0.1431 | 0.8232 | 0.8449 | 0.8339 | 0.9628 |
78
 
79
 
80
  ### Framework versions
 
82
  - Transformers 4.27.4
83
  - Pytorch 2.0.0+cu118
84
  - Datasets 2.11.0
85
+ - Tokenizers 0.13.3