Even though emails are identified as a significant source of abusive behaviors, little effort has been spent to date on the detection and prevention of such abusive behaviors. Through a mixed-method user study consisting of formative interviews and a survey, I explore abusive behavior and different types of abuse detection systems for emailing platforms in the context of Bangladesh. Inspired by those findings, following an iterative approach, a fully automated abuse detection system "Citadel" for emails has been developed. Evaluation results portray efficacy, efficiency, and user acceptance of "Citadel" in detecting and preventing abusive emails.
Even though emails are identified as a significant source of abusive behaviors, little effort has been spent to date on the detection and prevention of such abusive behaviors. Through a mixed-method user study consisting of formative interviews and a survey, I explore abusive behavior and different types of abuse detection systems for emailing platforms in the context of Bangladesh. Inspired by those findings, following an iterative approach, a fully automated abuse detection system "Citadel" for emails has been developed. Evaluation results portray efficacy, efficiency, and user acceptance of "Citadel" in detecting and preventing abusive emails.
Non-profit organizations play a critical role in welfare of street children, but the lack of proper interactions with people causes difficulty for small-scale organizations. Through mixed methods research, I studied limitations in existing interactions from the perspective of organizers, donors and general people, and explore possible technological interventions.
Paper published in COMPASS 2023, Poster published in CSCW Companion 2022
I explored utilizing a novel consumer behavior framework called Behavioral Reasoning Theory (BRT) to investigate consumers’ intentions to engage in circular consumption behavior. To validate the study model and test the research hypotheses proposed, a questionnaire-based survey was conducted with 293 respondents.
In genome sequencing, short reads have less error rates than long reads but is prone to repetitive sequences. I proposed an approach to combine best of both technologies and develop a maximum likelihood based model to estimate long read error parameters, extend short read contigs by using overlapping long reads and estimate gap lengths to create scaffolds between contigs.