Skip to Main content Skip to Navigation
Master Thesis

Text analysis and document classification of scientific articles related to herb-drug interactions

Abstract : Herb-drug interactions occur on simultaneous administration of herb with a therapeutic dose of the drug leading to potential health risks. Due to the perception of anything ‘natural’ ensuring safety or it being considered beneficial, there has been a rise in the use of herbal medicines in Western countries. revious work focuses mostly on drug-drug interactions, addressing the task of detecting interactions with a supervised approach using a set of lexical and synthetic features with promising results [1]. In this project, we perform unsupervised topic modelling on scientific abstracts to determine if a latent dirichlet allocation (LDA) model can provide us with terms that can help tag our model topics. The keywords obtained for the dominant topics of the model are useful to determine the tags or information on the herb and drug involved in the interactions. We also perform and compare various supervised classification models related to traditional machine learning and deep learning algorithms to classify the abstracts according to the level of severity of herb-drug interactions. To achieve this, a pipeline for data collection, text pre-processing, feature extraction, modelling and evaluation of model performance was conducted. Our results showed SVM classifier to have achieved high-quality performance with the highest F1-score of 0.86 compared to the other models.
Document type :
Master Thesis
Complete list of metadata
Contributor : Isped Univ Bordeaux Connect in order to contact the contributor
Submitted on : Tuesday, February 23, 2021 - 1:42:34 PM
Last modification on : Friday, April 30, 2021 - 6:43:50 PM
Long-term archiving on: : Monday, May 24, 2021 - 8:14:37 PM


  • HAL Id : dumas-03149905, version 1



Sneha Keerthi Nama Ravi. Text analysis and document classification of scientific articles related to herb-drug interactions. Santé publique et épidémiologie. 2020. ⟨dumas-03149905⟩



Record views


Files downloads