I want to fine-tune HunFlair-Gene on the JNLPBA dataset, but I am getting mostly null scores during training.
I believe this is because the dataset does not contain gene mentions but rather mentions of proteins and other labels. I tested my code on a dataset that contains only gene labels (BC2GM) and obtained good results.
To solve my problem, I tried loading the JNLPBA dataset from flair.datasets using:
from flair.datasets import JNLPBA
Load the JNLPBA dataset
jnlpba_dataset = JNLPBA()
Then, I attempted to rename the protein labels to "Gene" and remove the other labels, but it appears that Flair does not provide the ability to rename labels in a dataset.
How can I work with JNLPBA to compare my scores with those of HunFlair?
Question
Hello,
I want to fine-tune HunFlair-Gene on the JNLPBA dataset, but I am getting mostly null scores during training.
I believe this is because the dataset does not contain gene mentions but rather mentions of proteins and other labels. I tested my code on a dataset that contains only gene labels (BC2GM) and obtained good results.
To solve my problem, I tried loading the JNLPBA dataset from flair.datasets using:
from flair.datasets import JNLPBA
Load the JNLPBA dataset
jnlpba_dataset = JNLPBA()
Then, I attempted to rename the protein labels to "Gene" and remove the other labels, but it appears that Flair does not provide the ability to rename labels in a dataset.
How can I work with JNLPBA to compare my scores with those of HunFlair?