newfull5 / NLP_task

NLP_task (wip)

Topic Classification

• model: roberta-base
• datasets: ag_news
• train_set: 120K -> train_set: 100K, valid_set: 20K
• test_Set: 7.6K
• label: World(0), Sports(1), Business(2), Sci/Tech(3)

from transformers import AutoModelForSequenceClassification, AutoTokenizer 

model = AutoModelForSequenceClassification.from_pretrained('dhtocks/Topic-Classification')
tokenizer = AutoTokenizer.from_pretrained('roberta-base') 

Named Entity Recognition

• model: roberta-base
• datasets: conll2003_noMISC
• train_set: 14K
• valid_set: 3.2K
• test_set: 3.4K
• label:
  • O (0)
  • B-PER (1)
  • I-PER (2)
  • B-ORG (3)
  • I-ORG (4)
  • B-LOC (5)
  • I-LOC (6)

from transformers import AutoModelForTokenClassification, AutoTokenizer

model = AutoModelForTokenClassification.from_pretrained('dhtocks/Named-Entity-Recognition')
tokenizer = AutoTokenizer.from_pretrained('roberta-base')