Text Classification with Python -- 2
Budget: $10 – $30 CAD
DEEP LEARNING
goal: to develop advanced models for text classification and predict the category of scientific research papers. There are 20 labels, each given a numerical id. The labels correspond to categorization into different
subject areas. For each paper, you are given an id, the title, and the abstract.
Steps
• Split your training data into train, validation, and test sections so that
you can evaluate your model before testing it on test.csv
• Data augmentation and regularization.
• Cross-validation and hyperparameter tuning.
– Learning rate decay.
– Early stopping on the validation set.
• Model ensemble.
Datasets
• train.csv: This file provides the training nodes (label, node id).
• test.csv: This file provides the test nodes (node id,).
• text.csv: This file provides the features, i.e. the paper title and abstract
(paper id, title, abstract)
• nodeid2paperid.csv: This file provides the map from nodes to papers
(node id, paperid).
• sample.csv: This file provides a sample submission file in the correct
format (node id, label).
Ps. I have done the code using Bert and now I am looking for someone who can do this with cybert.
If you don't know what cybert is check this out https://medium.com/rapids-ai/cybert-28b35a4c81c4.
I will send you my code so it would be easier for you to apply cyBERT
goal: to develop advanced models for text classification and predict the category of scientific research papers. There are 20 labels, each given a numerical id. The labels correspond to categorization into different
subject areas. For each paper, you are given an id, the title, and the abstract.
Steps
• Split your training data into train, validation, and test sections so that
you can evaluate your model before testing it on test.csv
• Data augmentation and regularization.
• Cross-validation and hyperparameter tuning.
– Learning rate decay.
– Early stopping on the validation set.
• Model ensemble.
Datasets
• train.csv: This file provides the training nodes (label, node id).
• test.csv: This file provides the test nodes (node id,).
• text.csv: This file provides the features, i.e. the paper title and abstract
(paper id, title, abstract)
• nodeid2paperid.csv: This file provides the map from nodes to papers
(node id, paperid).
• sample.csv: This file provides a sample submission file in the correct
format (node id, label).
Ps. I have done the code using Bert and now I am looking for someone who can do this with cybert.
If you don't know what cybert is check this out https://medium.com/rapids-ai/cybert-28b35a4c81c4.
I will send you my code so it would be easier for you to apply cyBERT