Depthwise Separable Convolutions with Deep Residual Convolutions

XceptionNet, Depthwise Separable Convolutions, Deep Residual, CNN, CIFAR-10

In this project, we propose an optimized Xception architecture tailored for edge devices, aiming for lightweight and efficient deployment. We incorporate the depthwise separable convolutions with deep residual convolutions of the Xception architecture to develop a small and efficient model for edge devices. The resultant architecture reduces parameters, memory usage, and computational load. The proposed architecture is evaluated on the CIFAR 10 object detection dataset. The evaluation result of our experiment also shows the proposed architecture is smaller in parameter size and requires less training time while outperforming Xception architecture performance.

Ensemble Language Models for Multilingual Sentiment Analysis

BERT multilingual, AraBERT, XLM-RoBERTa, Instructions

In this project, I mainly explore sentiment analysis on tweet texts from SemEval-17 and the Arabic Sentiment Tweet dataset (ASTD). Moreover, I investigated four pretrained language models and proposed two ensemble language models. The findings include monolingual models exhibiting superior performance and ensemble models outperforming the baseline while the majority voting ensemble outperforms the English language.

Multiplatform Bangla Sentiment Analysis

Dataset, Transformers, LLMs, Instructions

The MUBASE dataset is a multiplatform dataset consisting of Tweets and Facebook posts, which are manually annotated with sentiment polarity. The annotation agreement of this manually annotated dataset shows an agreement score of 0.84, indicating a perfect agreement among the annotators.

MEDIC: a multi-task learning dataset for disaster image classification

Dataset, ResNet, VGG, EfficientNet, SqueezeNet, DenseNet

The MEDIC is the largest multi-task learning disaster related dataset, which is an extended version of the crisis image benchmark dataset. It consists data from several data sources such as CrisisMMD, data from AIDR and Damage Multimodal Dataset (DMD). The dataset contains 71,198 images.

Resources for Bangla Natural Language Processing (BanglaNLP)

Dataset, Transformers, BiLSTM, LMs

In our work A Review of Bangla Natural Language Processing Tasks and the Utility of Transformer Models, we provide a review of Bangla NLP tasks, resources, and tools available to the research community; we benchmark datasets collected from various platforms for nine NLP tasks using current state-of-the-art algorithms (i.e., transformer-based models). We provide comparative results for the studied NLP tasks by comparing monolingual vs. multilingual models of varying sizes. We report our results using both individual and consolidated datasets and provide data splits for future research. We reviewed a total of 108 papers and conducted 175 sets of experiments. Our results show promising performance using transformer-based models while highlighting the trade-off with computational costs. We hope that such a comprehensive survey will motivate the community to build on and further advance the research on Bangla NLP.

AmaderCAT

Language: PHP, JavaScript
Framework: CodeIgniter, JQuery, Bootstrap
Database:MySQL

The application AmaderCAT is the abbreviation of Amader Computer-assisted Translation. This application is developed for the purpose of building parallel corpus for Machine Translation system. The application contains a Translation Memory and a Glossary suggestions implementation that used for helping translators by providing TM and glossary suggestions. The application is collaborative and highly configurable for the translation task. It has the mechanism for crowd translation. You can use it as single user or a group/team. In future, we will add Machine Translation System in our application using Neural Network technologies.

Reviewer (Notable)

2024

2023

2022

2021

Professional Development