search

dmis-lab / biobert

Bioinformatics'2020: BioBERT: a pre-trained biomedical language representation model for biomedical text mining
http://doi.org/10.1093/bioinformatics/btz682
Other
1.92k stars 452 forks source link

Can biopert just predict the realtion using pertained models and not train as given in Biobert formula #99

Open dhwani2410 opened 4 years ago

dhwani2410 commented 4 years ago

I have two questions related to this:-

python run_re.py --task_name=$TASK_NAME --do_train=true --do_eval=true --do_predict=true --vocab_file=$BIOBERT_DIR/vocab.txt --bert_config_file=$BIOBERT_DIR/bert_config.json --init_checkpoint=$BIOBERT_DIR/model.ckpt-1000000 --max_seq_length=128 --train_batch_size=32 --learning_rate=2e-5 --num_train_epochs=3.0 --do_lower_case=false --data_dir=$RE_DIR --output_dir=$OUTPUT_DIR

1) Can we run the prediction and not train. For the same how should the input file be made? As the input file to train consist of three columns 1)Index 2)Sentences with (NER replaced with a keyword) 3) Labels. Since I want to run in an unknown dataset and I don't have labels, can just Sentence column be the input to model?

2) What kind of a relation does the model tell? Is it positive or negative or just that there exists some relationship between the NER present in the sentences(for example GENES and DISEASES)

ghost commented 3 years ago

@dhwani2410, I have same question with you, have you got some answer regarding this?