arXiv subjects

Ipsita Mohanty

Publications and source records attributed to Ipsita Mohanty.

5 recordsLinked to original sources

arXiv2023

DEFTri: A Few-Shot Label Fused Contextual Representation Learning For Product Defect Triage in e-Commerce

Defect Triage is a time-sensitive and critical process in a large-scale agile software development lifecycle for e-commerce. Inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams. This work proposes a novel framework for automated defect triage (DEFTri) using fine-tuned state-of-the-art pre-trained BERT on labels fused text embeddings to improve contextual representations from human-generated product defects. For our multi-label text classification defect triage task, we also introduce a Walmart proprietary dataset of product defects using weak supervision and adversarial learning, in a few-shot setting.

cs.SE↗

arXiv2023

Comprehensive and Comprehensible Data Catalogs: The What, Who, Where, When, Why, and How of Metadata Management

Data management tasks require access to metadata, which is increasingly tracked by databases called data catalogs. Current catalogs are too dependent on users' understanding of data, leading to difficulties in large organizations of users with different skills: catalogs either make metadata easy for users to store and difficult to retrieve, or they make it easy to retrieve, but difficult to store. In this paper, we present 5W1H+R, a new catalog mental model that is comprehensive in the metadata it represents, and comprehensible in that it permits all users to locate metadata easily. We demonstrate these properties via a user study. We then discuss practical guidelines for implementing the new mental model. We conclude mental models are important to make data catalogs more useful and to boost metadata management efforts.

cs.DB↗

arXiv2021

Emotions are Subtle: Learning Sentiment Based Text Representations Using Contrastive Learning

Contrastive learning techniques have been widely used in the field of computer vision as a means of augmenting datasets. In this paper, we extend the use of these contrastive learning embeddings to sentiment analysis tasks and demonstrate that fine-tuning on these embeddings provides an improvement over fine-tuning on BERT-based embeddings to achieve higher benchmarks on the task of sentiment analysis when evaluated on the DynaSent dataset. We also explore how our fine-tuned models perform on cross-domain benchmark datasets. Additionally, we explore upsampling techniques to achieve a more balanced class distribution to make further improvements on our benchmark tasks.

cs.CL↗

arXiv2012

Information Retrieval From Internet Applications For Digital Forensic

Advanced internet technologies providing services like e-mail, social networking, online banking, online shopping etc., have made day-to-day activities simple and convenient. Increasing dependency on the internet, convenience, and decreasing cost of electronic devices have resulted in frequent use of online services. However, increased indulgence over the internet has also accelerated the pace of digital crimes. The increase in number and complexity of digital crimes has caught the attention of forensic investigators. The Digital Investigators are faced with the challenge of gathering accurate digital evidence from as many sources as possible. In this paper, an attempt was made to recover digital evidence from a system's RAM in the form of information about the most recent browsing session of the user. Four different applications were chosen and the experiment was conducted across two browsers. It was found that crucial information about the target user such as, user name, passwords, etc., was recoverable.

cs.CR↗

arXiv2010

2-Variable Boolean Operation -- its use in Pattern Formation

In this paper the theory of 2-Variable Boolean Operation (2-VBO) has been discussed on a pair of n-bit strings. 2-VBO serves to bring out the relation between numbers which when plot on a 2-D surface form interesting patterns; patterns that may be fixed, periodic, chaotic or complex. Some of these patterns represent natural fractals. This paper also provides mathematical analysis corresponding to each of the obtained patterns, which would aid to understanding their formation. 2-VBO is an attempt towards the production and classification of patterns which represent various mathematical models and naturally occurring phenomena.

nlin.CD↗