Rupak Kumar Das - Research and Projects

Publications

Koujalgi, S., Anderson, A., Adenuga, I., Soneji, S., Dikkala, R., Nader, T.G., Soccio, L., Panda, S., Das, R.K., Burnett, M. and Dodge, J., 2024. How to Measure Human-AI Prediction Accuracy in Explainable AI Systems. arXiv preprint arXiv:2409.00069.
Dodge, Jonathan, Antoinette Diawuo, Enyan Dai, and Rupak Das. "Learning About Explainable AI with Very Little Programming: For 4 th Year Undergrads (or Younger)." In Innovative Practices in Teaching Information Sciences and Technology: Further Experience Reports and Reflections, pp. 121-138. Cham: Springer Nature Switzerland, 2024.
Islam, Mahbubul, Hossain Md. Mubashshir Jamil, Samiul Ahsan Pranto, Rupak Kumar Das, Al Amin, and Arshia Khan. 2024. "Future Industrial Applications: Exploring LPWAN-Driven IoT Protocols" Sensors 24, no. 8: 2509. https://doi.org/10.3390/s24082509
Khan, A. A., Das, R. K., Martin, A., Dowling, D., & Imtiaz, R. (2024). Creating an Intelligent Dementia-Friendly Living Space: A Feasibility Study Integrating Assistive Robotics, Wearable Sensors, and Spatial Technology. arXiv preprint arXiv:2401.00959.
Das, R.K.; Martin, A.; Zurales, T.; Dowling, D.; Khan, A. A Survey on EEG Data Analysis Software. Sci 2023, 5, 23. https:// doi.org/10.3390/sci5020023
Das, R.K.; Imtiaz, N.Z.; Khan, A. Toward Affirmation of Recovery of Deeply Embedded Autobiographical Memory with Background Music and Identification of an EEG Biomarker in Combination with EDA Signal Using Wearable Sensors. Clin. Transl. Neurosci. 2022, 6, 26
Lee. A.; Mhatre.J.; Das. R.K.; Hong. M. Hybrid Mobile Cloud Computing Architecture with Load Balancing for Healthcare Systems. Computers, Materials & Continua 2023, 74(1), 435-452
Das, R. K., & Lee, A. (2020, October). A Study of Load-Balancing Solutions of Mobile Cloud Computing for Next-Generation Mobile Applications. In Proceedings of the International Conference on Research in Adaptive and Convergent Systems (pp. 119-123)
Hosen, M. R., Hasan, S., Hasan, M. M., & Das, R. K. (2015, May). Age classification based on EMG signal using Artificial Neural Network. In 2015 International Conference on Electrical Engineering and Information Communication Technology (ICEEICT) (pp. 1-5). IEEE
Das, R. K. (2022). Towards affirmation of recovery of deeply embedded autobiographical memory and identification of an EEG biomarker using wearable sensors (Doctoral dissertation, UNIVERSITY OF MINNESOTA).
Elderly Friendly exercises: A PRISMA-guided review of the literature to enable awareness of exercises that are recommended for the elderly (Submitted)
Performance Analysis on GPU-Backed Data Analytics and Deep Learning (Submitted)
Performance Analysis Between Machine Learning Classifiers for Fake News Detection (Submitted)

Research

Design of an Intelligent Tutoring System: The primary objective of the vision paper is to demonstrate an end-to-end framework that can identify the gap between students’ prior knowledge and lack of understanding of a topic. Based on that, we aim to develop an intelligent tutoring system to help students by recommending appropriate materials with suitable and personalized learning styles.
Media bias on Twitter by media outlets during the Russian Invasion of Ukraine in 2022: In this project, we use different methodologies to discover the bias shown by different news media outlets on Twitter towards any specific entity or topic during the Russian Invasion of Ukraine in 2022.
Identification of an EEG Biomarker for memory recall using EEG and EDA with music: This project aimed to find an EEG biomarker to verify memory recall. Forty participants participated in this experiment, and Electroencephalogram (EEG) and Electrodermal Activity (EDA) data were collected while playing background music.
Design of a Dementia friendly environment: We investigated the feasibility of combining assistive therapeutic robots, an intelligent pill dispenser, and wearable sensors to complement human care-giving in a space equipped with sensors and a dementia-friendly atmosphere.
Feature Extraction and Classification of EEG signals for body movement in different directions: This research was designed to develop a machine-learning model to classify body movement in different directions using features extracted from Electroencephalogram (EEG) signals collected from six participants. A Support Vector Machine (SVM) classifier was used to classify the body movement.

Projects

Performance Analysis Between Machine Learning Classifiers for Fake News Detection

In this project, different feature engineering methods like TF-IDF, count vector, and word embedding were used to generate features. Those features are used with Support Vector Machine (SVM), Logistic Regressions, Random Forest, Naive Bayes Classifier, and Decision tree classifier as supervised machine learning models, and Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) as the deep learning models. A publicly available dataset with 10,302 fake news articles and 10,379 real news articles was collected from Kaggle for this project. The final results show that Deep learning algorithms perform slightly better than supervised learning algorithms in detecting fake news.

Classification Metrics with TF-IDF Vectorizer

Classification Metrics with Count Vectorizer

Classification Metrics with Word Embedding

Analyzing the Impact of AI Writing Assistants on Writing Proficiency: Does Usage Enhance or Hinder Independent Writing Skills?

In this project, through human participant research, we figured out the possible enhancing and hindering effects of using writing assistants. Writing assistants can improve lexical richness, readability, and overall writing performance. On the other hand, it may sometimes provide wrong and inaccurate suggestions. Also, writers may lose their unique writing style and originality.

Impact of writing assistant on Type Token Ratio (TTR)

Impact of writing assistant on Lexical Frequency Score (LFS)

Impact of writing assistant on readability

Impact of writing assistant on overall score

Twitter Sentiment Analysis (SemEval2017) using BERT

This project is part of Task 4 of SemEval 2017. It involves Twitter sentiment analysis, automatically classifying tweets' sentiments into three categories: positive, negative, and neutral. A transformer-based architecture named BERT is used for this classification. For a baseline proof-of-concept model, I used a Naive Bayes Classifier, which is often a good choice as a baseline model. The Code for this project can be found here.

Design and Implementation of a Database System

This is a class project, of course, CS 5322. The main objectives were to design and implement a database system based on the given scenario. Finally, a Java database application was implemented for some queries. The Code for this project can be found here.

SuperGlue Tasks (RTE and BoolQ) using BERT

This is a class project of CS 5642 (Ad. NLP). I worked on two SuperGlue tasks in this project: Recognizing Textual Entailment (RTE) and BoolQ.

RTE: This task involves determining whether a given pair of sentences entail each other, contradict each other, or are neutral.

BoolQ: This task involves answering yes or no questions based on a given passage.

The code for this project can be found here.

Design of a seq2seq Machine Translator

Machine Translation is the task of translating a sentence from one language to another using a Neural Machine Translator (NMT). It is a way to do Machine Translation with a single neural network. This architecture, called sequence-to-sequence (or seq2seq), involves two RNNs. Bilingual Evaluation Understudy (BLEU) is used to evaluate the designed NMT. The code for this project can be found here.

Design of Neural Language Mode

A Neural Language Model (NLM) is an artificial neural network designed to predict the probability of the next word in a sequence of words given the previous words. It is a statistical language model trained on large amounts of text data. It can be used to generate new sentences or evaluate the likelihood of a given sequence of words. The code for this project can be found here.

Career Counselor ChatBot Design

This is a CS 5242 class project. The main objective was to design a rule-based career counselor chatbot for college students near graduation.

The code for this project can be found here.

An Ngram language model from scratch

This is a Language model that generates text based on Unigram, Bigram, and Trigram models without any NLP-based library. Training data of over 1,000,000 tokens have been used to train this model. Data was collected from Project Gutenberg. The code for this project can be found here.

Sentiment Analysis using decision list from scratch

This is a sentiment classifier trained for classifying movie reviews. The objective is to use a decision list tree to classify test reviews using unigram, bigram, and not handling features and find the accuracy, precision, and recall. The code for this project can be found here.

Word-by-word co-occurrence matrix from scratch

This program creates a word-by-word co-occurrence matrix for a corpus. The cell values in this matrix should be the Pointwise Mutual Information (PMI) between 2 words. The code for this project can be found here.

Part of speech tagging from scratch

The main objective of this project is to design a part-of-speech tagger, a software tool that analyzes text and assigns a part-of-speech tag to each word in the text based on its grammatical role and context within the sentence. The part-of-speech tags indicate the syntactic category of the word, such as noun, verb, adjective, adverb, pronoun, preposition, conjunction, or interjection. The code for this project can be found here.

Machine Reading Comprehension

The main objective of this project is to design a Machine Reading Comprehension that involves extracting relevant information from the passage and generating a response that answers a given question. The code for this project can be found here.

Live Face Mask Detection Using CNN

Live face mask detection using Convolutional Neural Networks (CNN) is an application of computer vision that aims to automatically detect whether a person is wearing a face mask in real-time. This application is particularly relevant in the context of the COVID-19 pandemic, where wearing face masks has become a critical measure for preventing the spread of the virus. The code for this project can be found here.

Implementation of Hadoop Multi-Node Cluster

A Hadoop Multi-Node Cluster is designed and implemented in this project for processing and analyzing large data sets. In a Hadoop multi-node cluster, data is stored and processed across multiple nodes to enable the parallel processing of large data sets. The cluster consists of a master node and multiple worker nodes, where the master node manages the distribution of tasks across the worker nodes. The code for this project can be found here.

Design and Implementation of Blood Glucose Meter
Design and Implementation of Smoke Detector

Page updated

Google Sites

Report abuse