Skip to content

Relation Extraction on MIMIC-III Data using TF-IDF, Bag-of-Words, Word2Vec, Spacy, BERT, and Sentence-BERT

Notifications You must be signed in to change notification settings

dhannywi/Relation_Extraction

Repository files navigation

Relation Extraction on MIMIC-III Data

AI in Health group project in collaboration with Dr. Yifan Peng from Cornell Medicine.

The research project focuses on exploring different possibilities to build a successful model that can classify the type of relation type available in a sentence of medical text. In our research, we utilized a variety of feature extraction methods such as TF-IDF, Bag-of-Words, Word2Vec, Spacy, BERT, and Sentence-BERT. For the classification task, we implemented models such as Decision Tree, Random Forest, Long Short Term Memory model et cetera. We found that in our smaller datasets, simpler models do surprisingly better than the complex model. We also found that SentenceBERT provides excellent representation which improved the results for almost all of the classification models.

Detailed description of the research process and findings are available in the PDF report and presentation slides.

Authors

Dhanny Indrakusuma
Redoan Rahman
Patrick Sui

About

Relation Extraction on MIMIC-III Data using TF-IDF, Bag-of-Words, Word2Vec, Spacy, BERT, and Sentence-BERT

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published