Training the biencoder

[anshks]

I was trying to reproduce the code. However, I am facing issues with training the biencoder. Could you please tell how to finetune the biencoder on the datasets mentioned in the paper? Also, can you provide the WikilinksNED Unseen-Mentions dataset which you used? @fabiopetroni

[ledw]

@anshks We will update with an instruction on how to fine-tune the model shortly. For the WikilinksNED Unseen-Mentions dataset, we obtained it from the authors of this paper: https://ojs.aaai.org/index.php/AAAI/article/view/6380

[nlp4whp]

@anshks We will update with an instruction on how to fine-tune the model shortly. For the WikilinksNED Unseen-Mentions dataset, we obtained it from the authors of this paper: https://ojs.aaai.org/index.php/AAAI/article/view/6380

+1 Waiting for fine-tuning note :)

[shzamanirad]

@anshks We will update with an instruction on how to fine-tune the model shortly. For the WikilinksNED Unseen-Mentions dataset, we obtained it from the authors of this paper: https://ojs.aaai.org/index.php/AAAI/article/view/6380

@ledw any progress on this? I'm trying to train the bi-encoder, but can't find the details of how you trained it (e.g. scripts, hyperparameters).

[abhinavkulkarni]

It is indeed possible to train a model on your own custom dataset. I would recommend first training a zeshel (Zero-shot learning) model, then bring your data in the same format as zeshel and train your model.

Scripts to download and pre-process zeshel data set are in the repository.