akdeniz27 commited on
Commit
13c7df7
1 Parent(s): dba453c

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +37 -0
README.md ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ --
2
+ language: tr
3
+ ---
4
+
5
+ # Turkish Named Entity Recognition (NER) Model
6
+
7
+ This model is the fine-tuned model of dbmdz/bert-base-turkish-cased
8
+ using a reviewed version of well known Turkish NER dataset
9
+ (https://github.com/stefan-it/turkish-bert/files/4558187/nerdata.txt).
10
+
11
+ # Fine-tuning parameters:
12
+ ```
13
+ task = "ner"
14
+ model_checkpoint = "dbmdz/bert-base-turkish-cased"
15
+ batch_size = 8
16
+ label_list = ['O', 'B-PER', 'I-PER', 'B-ORG', 'I-ORG', 'B-LOC', 'I-LOC']
17
+ max_length = 512
18
+ learning_rate = 2e-5
19
+ num_train_epochs = 3
20
+ weight_decay = 0.01
21
+ ```
22
+
23
+ # How to use:
24
+ ```
25
+ model = AutoModelForTokenClassification.from_pretrained("akdeniz27/bert-base-turkish-cased-ner")
26
+ tokenizer = AutoTokenizer.from_pretrained("akdeniz27/bert-base-turkish-cased-ner")
27
+ ner = pipeline('ner', model=model, tokenizer=tokenizer, aggregation_strategy="first")
28
+ NER("text")
29
+
30
+ # Pls refer "https://huggingface.co/transformers/_modules/transformers/pipelines/token_classification.html" for entity grouping with aggregation_strategy parameter.
31
+ ```
32
+
33
+ # Reference test results:
34
+ * accuracy: 0.9933935699477056
35
+ * f1: 0.9592969472710453
36
+ * precision: 0.9543530277931161
37
+ * recall: 0.9642923563325274