Add Bert Model

[sam5658]

Prerequisites

• [X] I am an Open Source Contributor
• [X] I checked to make sure that this issue has not already been filed

Description

Hi I'm Shubham | Contributor GSSOC'23. Many models are available for performing sentiment analysis(in our case, patient review), which has proven to cause a significant rise in prediction accuracy. Therefore I'm willing to implement the BERT model which can be utilized with the help of the hugging face "transformer" library.

Implementation:

• Deep Learning Models such as Bert are known for their great Transformer based Architecture, Bert's architecture needs to be fine-tuned on our Drug Dataset.
• hugging face transformers have various functionalities which can ease our model training keeping the code readable.
• I suggest training the model with the help of Pytorch library(only for creating CustomDataset) and DataLoaders
• All of these points are needed to be coded in a different file than "main.py" so that those unfamiliar with the PyTorch library can work still go through the code base.

Deliverables:

• A separate file from "main.py" implementing and training the BERT model for improving the accuracy.
• The deep learning model can then be imported into the file "main.py" with just a few lines of code.

issue Further for visualisation, I suggest using seaborn only as sklearn.metrics.plot_confusion_matrics is deprecated.

Screenshots

No response

Code of Conduct

• [X] I follow Contributing Guidelines of this project.

[Rakesh9100]

@sam5658 Okay Assigning you, Start working

[Rakesh9100]

@sam5658 Status of your work

[sam5658]

The data set was quite big, that's why I couldn't train the model as it was reading the entire CPU ram, but the code has been optimized now and I will put the model in training and give a pull request in a couple of days.

On Sat, 27 May 2023 at 03:07, Rakesh Roshan @.***> wrote:

@sam5658 https://github.com/sam5658 Status of your work

— Reply to this email directly, view it on GitHub https://github.com/Rakesh9100/ML-Project-Drug-Review-Dataset/issues/20#issuecomment-1564985481, or unsubscribe https://github.com/notifications/unsubscribe-auth/AYGB77G6CRNVTLCSNDEMJILXIEPALANCNFSM6AAAAAAYJJU7RU . You are receiving this because you were mentioned.Message ID: @.*** com>

[Rakesh9100]

@sam5658 Update pls

[Rakesh9100]

Closing for having no response.