Text Classification with pre-trained Embedding

Problem

Classify given set of Pubmed abstracts (biomedical literature abstracts) into four classes:

• Abstracts containing Drug adverse events
• Abstracts containing Congenital anomalies
• Abstracts containing both (a) and (b)
• Others

Dataset: Pubmed (https://pubmed.ncbi.nlm.nih.gov/)

Required Libraries

• python 3
• numpy
• tenforflow
• keras
• sklearn
• pandas
• bs4
• requests
• matplotlib
• scipy

Download Data

Code data_download.py will downlod all the required data in four classes

• Each class includes 700 examples
• Class other has two time more examples (1400) to keep all classes ballanced

$ python data_download.py

Train the model

To train the model run the following

$ python NLP_Classification.py --task train

To evaluate the model performance the the following

$ python NLP_Classification.py --task test

Conclution

There are a large number of possibility to train such a model, however the following are important to mention

• Generally Neural Netwrok based models performes better
• First a tokenizer generates arrays from text
• Second an Embedding layer generates array representation for each sequesnce
• WE can use RNN (simple or LSTM), CNN or attention based models
• My results is based on a CNN model