Project descriptions
This project delves into the fundamental aspects of natural language processing (NLP), which involves the processing of text data to make predictions. For this purpose, I utilized the PubMed200kRCT_medical_abstracts dataset, which consists of approximately 200,000 labeled Randomized Controlled Trial (RCT) abstracts, as presented in the paper "PubMed 200k RCT: a Dataset for Sequential Sentence Classification in Medical Abstracts" published in 2017. For further in-depth information about the dataset, I recommend referring to the original paper for comprehensive details.
In NLP, there are various approaches to handle text data prior to training, such as text/word level or character level. Text level involves treating each word as a token with a unique ID, whereas character level entails assigning a unique ID to each character and training the model accordingly. In this project, I have covered both approaches with a simple model architecture for demonstration purposes. Despite the simplicity, the accuracy of the model is noteworthy. Feel free to check out the project for more details!