Article: Extraction of drug-drug interaction information using a deep neural network Journal: International Journal of Data Mining and Bioinformatics (IJDMB) 2021 Vol.25 No.3/4 pp.181 - 200 Abstract: The information about Drug-Drug Interaction (DDI) is available in biomedical literature and extraction of this information manually is an extremely challenging and arduous task. DDI information helps medical practitioners to suggest various combination of drugs. The existing DDI extraction systems, based on traditional methods like SVM, require manually defined features, which is a time consuming task. Hence, deep learning algorithms, which eliminate the need for manual feature engineering, are applied for classification of sentences in biomedical journals. In this paper, a 2-Input Layer BiLSTM-based DDI extraction model is developed to assist health care workers to find out the DDI information by themselves. This model takes two features as input, the word embeddings applied to words in a sentence and embeddings for target drug pair. The proposed 2-Input Layer BiLSTM model has outperformed existing models by an F-Score of 4% for binary classification and 10% for multiclass classification. Inderscience Publishers - linking academia, business and industry through research

Title: Extraction of drug-drug interaction information using a deep neural network

Authors: Serena Rajakumar; G. Kavitha; I. Sathik Ali

Addresses: Department of Information Technology, B.S.A. Crescent Institute of Science & Technology, Chennai, Tamil Nadu, India ' Department of Information Technology, B.S.A. Crescent Institute of Science & Technology, Chennai, Tamil Nadu, India ' Department of Information Technology, B.S.A. Crescent Institute of Science & Technology, Chennai, Tamil Nadu, India

Abstract: The information about Drug-Drug Interaction (DDI) is available in biomedical literature and extraction of this information manually is an extremely challenging and arduous task. DDI information helps medical practitioners to suggest various combination of drugs. The existing DDI extraction systems, based on traditional methods like SVM, require manually defined features, which is a time consuming task. Hence, deep learning algorithms, which eliminate the need for manual feature engineering, are applied for classification of sentences in biomedical journals. In this paper, a 2-Input Layer BiLSTM-based DDI extraction model is developed to assist health care workers to find out the DDI information by themselves. This model takes two features as input, the word embeddings applied to words in a sentence and embeddings for target drug pair. The proposed 2-Input Layer BiLSTM model has outperformed existing models by an F-Score of 4% for binary classification and 10% for multiclass classification.

Keywords: recurrent neural network; natural language processing; sentence classification; information extraction; knowledge extraction; binary classification; multiclass classification; bidirectional LSTM; word embedding; GloVe embedding.

DOI: 10.1504/IJDMB.2021.122855

International Journal of Data Mining and Bioinformatics, 2021 Vol.25 No.3/4, pp.181 - 200

Received: 09 Aug 2020
Accepted: 24 Nov 2021
Published online: 13 May 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Extraction of drug-drug interaction information using a deep neural network

Keep up-to-date