Soham Dasgupta

Soham Dasgupta

Computer Science and Business Finance Undergraduate

Publications

“Cybersecurity Knowledge Graph Improvement with Graph Neural Networks”

Published in BigCyber Conference @ IEEE Big Data 2021, 2021

Cybersecurity Knowledge Graphs (CKGs) help in aggregating information about cyber-events. CKGs have assertions in the form of semantic triples. The triples describe a relationship between a subject and object, both of which are cybersecurity entities. In this paper, we describe a novel method that associates a score with the semantic triples asserted in the CKG using deep learning. We use semantic triples that we know are correct, in a supervised machine learning algorithm that produces the output for each relationship. In particular, we use Graph Convolutional Neural Networks (GCN) on a dataset of CKGs that can be used to ascertain the scores for each semantic triple.

Download here

“Adverse Drug Event Prediction using Noisy Literature-Derived Knowledge Graphs”

Published in Journal of Medical Internet Research(JMIR), 2021

Adverse drug events (ADEs) are unintended side effects of drugs that cause substantial clinical and economic burdens globally. A wealth of information, which facilitates ADE discovery, lies in the growing body of biomedical literature. Knowledge graphs (KGs) encode this information from the literature, where the vertices and the edges represent clinical concepts and their relations, respectively. In this paper, we use confidence scores during representation learning, so that the learned embeddings yield better features for ADE prediction models.

Download here

“Understanding the role of affect dimensions in detecting emotions from tweets”

Published in 44th International ACM SIGIR Conference on Research and Development in Information Retrieval 2021, 2021

In this work, we make an attempt to bridge this gap by proposing a Transformers based multi-task framework that jointly trains multi-label emotion classification and multi-dimensional emotion regression, thereby exploiting the inter-relatedness between the two tasks for learning better shared representations for subjectivity analysis. We demonstrate the efficacy of our architecture on two benchmark datasets, EMOBANK and Affect in Tweets. We were further motivated by the need to develop a more efficient end-to-end deployable solution for better public health monitoring from social media conversations, which has become more important than ever owing to the tremendous detrimental impacts on the physical and mental well-being of the world population caused by the ongoing COVID-19 pandemic.

Download here

“The Role of a Mask in the performance of Deep Neural Networks to segment cell nuclei from microscopy images”

Published in Journal of Emerging Investigators, 2021

Cell segmentation is the task of identifying cell nuclei instances in fluorescence microscopy images. It plays a key role in biomedical analysis tasks like cell characterization, cancer cell identification, and gene expression measurement. The goal of this paper is to benchmark the performance of representative deep learning techniques for cell nuclei segmentation using standard datasets and common evaluation criteria. We investigate whether for cell nuclei segmentation, learning the nuclei masks in parallel with boundary detection features of the image provides superior performance and significantly cleaner separation of nuclei from images. We further explore the trade-off between increased accuracy, achieved through more complex deep learning models, and the heavy requirements imposed on both computational resources and training times. We believe this paper establishes an important baseline for cell nuclei segmentation, enabling researchers to continually refine and deploy neural models for real-world clinical applications.

Download here

““A Comparative Study of Deep Learning based Named Entity Recognition Algorithms for Cybersecurity””

Published in IEEE Big Data 2020 Conference, 2020

Named Entity Recognition (NER) is important in the cybersecurity domain. It helps researchers extract cyber threat information from unstructured text sources. The extracted cyber-entities or key expressions can be used to model a cyber-attack described in an open-source text. A large number of general-purpose NER algorithms have been published that work well in text analysis. In this paper, we compare the recent deep learning-based NER algorithms on a cybersecurity dataset. We created a cybersecurity dataset collected from various sources, including “Microsoft Security Bulletin” and “Adobe Security Updates”. Some of these approaches proposed in literature were not used for Cybersecurity. Others are innovations proposed by us. This comparative study helps us identify the NER algorithms that are robust and can work well in sentences taken from a large number of cybersecurity sources. We tabulate their performance on the test set and identify the best NER algorithm for a cybersecurity corpus. We also discuss the different embedding strategies that aid in the process of NER for the chosen deep learning algorithms.

Download here