arXiv subjects

Gouri Ginde

Publications and source records attributed to Gouri Ginde.

25 records · Page 2Linked to original sources

arXiv2024

Trustworthy and Responsible AI for Human-Centric Autonomous Decision-Making Systems

Artificial Intelligence (AI) has paved the way for revolutionary decision-making processes, which if harnessed appropriately, can contribute to advancements in various sectors, from healthcare to economics. However, its black box nature presents significant ethical challenges related to bias and transparency. AI applications are hugely impacted by biases, presenting inconsistent and unreliable findings, leading to significant costs and consequences, highlighting and perpetuating inequalities and unequal access to resources. Hence, developing safe, reliable, ethical, and Trustworthy AI systems is essential. Our team of researchers working with Trustworthy and Responsible AI, part of the Transdisciplinary Scholarship Initiative within the University of Calgary, conducts research on Trustworthy and Responsible AI, including fairness, bias mitigation, reproducibility, generalization, interpretability, and authenticity. In this paper, we review and discuss the intricacies of AI biases, definitions, methods of detection and mitigation, and metrics for evaluating bias. We also discuss open challenges with regard to the trustworthiness and widespread application of AI across diverse domains of human-centric decision making, as well as guidelines to foster Responsible and Trustworthy AI models.

cs.AI↗

arXiv2024

AROhI: An Interactive Tool for Estimating ROI of Data Analytics

The cost of adopting new technology is rarely analyzed and discussed, while it is vital for many software companies worldwide. Thus, it is crucial to consider Return On Investment (ROI) when performing data analytics. Decisions on "How much analytics is needed"? are hard to answer. ROI could guide decision support on the What?, How?, and How Much? Analytics for a given problem. This work details a comprehensive tool that provides conventional and advanced ML approaches for demonstration using requirements dependency extraction and their ROI analysis as use case. Utilizing advanced ML techniques such as Active Learning, Transfer Learning and primitive Large language model: BERT (Bidirectional Encoder Representations from Transformers) as its various components for automating dependency extraction, the tool outcomes demonstrate a mechanism to compute the ROI of ML algorithms to present a clear picture of trade-offs between the cost and benefits of a technology investment.

cs.SE↗

arXiv2024

Towards Extracting Ethical Concerns-related Software Requirements from App Reviews

As mobile applications become increasingly integral to our daily lives, concerns about ethics have grown drastically. Users share their experiences, report bugs, and request new features in application reviews, often highlighting safety, privacy, and accountability concerns. Approaches using machine learning techniques have been used in the past to identify these ethical concerns. However, understanding the underlying reasons behind them and extracting requirements that could address these concerns is crucial for safer software solution development. Thus, we propose a novel approach that leverages a knowledge graph (KG) model to extract software requirements from app reviews, capturing contextual data related to ethical concerns. Our framework consists of three main components: developing an ontology with relevant entities and relations, extracting key entities from app reviews, and creating connections between them. This study analyzes app reviews of the Uber mobile application (a popular taxi/ride app) and presents the preliminary results from the proposed solution. Initial results show that KG can effectively capture contextual data related to software ethical concerns, the underlying reasons behind these concerns, and the corresponding potential requirements.

cs.SE↗

arXiv2024

PRAGyan -- Connecting the Dots in Tweets

As social media platforms grow, understanding the underlying reasons behind events and statements becomes crucial for businesses, policymakers, and researchers. This research explores the integration of Knowledge Graphs (KGs) with Large Language Models (LLMs) to perform causal analysis of tweets dataset. The LLM aided analysis techniques often lack depth in uncovering the causes driving observed effects. By leveraging KGs and LLMs, which encode rich semantic relationships and temporal information, this study aims to uncover the complex interplay of factors influencing causal dynamics and compare the results obtained using GPT-3.5 Turbo. We employ a Retrieval-Augmented Generation (RAG) model, utilizing a KG stored in a Neo4j (a.k.a PRAGyan) data format, to retrieve relevant context for causal reasoning. Our approach demonstrates that the KG-enhanced LLM RAG can provide improved results when compared to the baseline LLM (GPT-3.5 Turbo) model as the source corpus increases in size. Our qualitative analysis highlights the advantages of combining KGs with LLMs for improved interpretability and actionable insights, facilitating informed decision-making across various domains. Whereas, quantitative analysis using metrics such as BLEU and cosine similarity show that our approach outperforms the baseline by 10\%.

cs.IR↗

arXiv2018

Use of NoSQL database and visualization techniques to analyze massive scholarly article data from journals

Visualization of the massive data is a challenging endeavor. Extracting data and providing graphical representations can aid in its effective utilization in terms of interpretation and knowledge discovery. Publishing research articles has become a way of life for academicians. The scholarly publications can shape-up the professional growth of authors and also expand the research and technological growth of a country, continent and other demographic regions. Scholarly articles have grown in gigantic numbers that are published in different domains by various journals. Information related to articles, authors, their affiliations, number of citations, country, publisher, references and other information is like a gold mine for statisticians and data analysts. This data when used skillfully, via visual analysis tool, can provide valuable understanding and can aid in deeper exposition for researchers working in domains like scientometrics and bibliometrics. Since the data is not readily available, we used Google scholar, a comprehensive and free repository of scholarly articles, as data source for our study. Data was scraped from Google scholar and stored as a graph and later visualized in the form of nodes and its relationships, which offered discerning and concealed information of growing impact of articles, journals and authors in their domains. Not only this, evident domain shift of an author, various research domains spread for an author, predicting emerging domain and subdomains, detecting cartel behavior at Journal and author-level was also depicted by graphical analysis. Neo4j graph database was used in the background to help store the data in structured manner.

cs.DL↗

arXiv2016

Visualisation of massive data from scholarly Article and Journal Database A Novel Scheme

Scholarly articles publishing and getting cited has become a way of life for academicians. These scholarly publications shape up the career growth of not only the authors but also of the country, continent and the technological domains. Author affiliations, country and other information of an author coupled with data analytics can provide useful and insightful results. However, massive and complete data is required to perform this research. Google scholar which is a comprehensive and free repository of scholarly articles has been used as a data source for this purpose. Data scraped from Google scholar when stored as a graph and visualized in the form of nodes and relationships, can offer discerning and concealed information. Such as, evident domain shift of an author, various research domains spread for an author, prediction of emerging domain and sub domains, detection of journal and author level citation cartel behaviors etc. The data from graph database is also used in computation of scholastic indicators for the journals. Eventually, econometric model, named Cobb Douglas model is used to compute the journals Modeling "Internationality" Index based on these scholastic indicators.

cs.HC↗

arXiv2016

ScientoBASE: A Framework and Model for Computing Scholastic Indicators of non-local influence of Journals via Native Data Acquisition algorithms

Defining and measuring internationality as a function of influence diffusion of scientific journals is an open problem. There exists no metric to rank journals based on the extent or scale of internationality. Measuring internationality is qualitative, vague, open to interpretation and is limited by vested interests. With the tremendous increase in the number of journals in various fields and the unflinching desire of academics across the globe to publish in "international" journals, it has become an absolute necessity to evaluate, rank and categorize journals based on internationality. Authors, in the current work have defined internationality as a measure of influence that transcends across geographic boundaries. There are concerns raised by the authors about unethical practices reflected in the process of journal publication whereby scholarly influence of a select few are artificially boosted, primarily by resorting to editorial maneuvres. To counter the impact of such tactics, authors have come up with a new method that defines and measures internationality by eliminating such local effects when computing the influence of journals. A new metric, Non-Local Influence Quotient(NLIQ) is proposed as one such parameter for internationality computation along with another novel metric, Other-Citation Quotient as the complement of the ratio of self-citation and total citation. In addition, SNIP and International Collaboration Ratio are used as two other parameters.

cs.DL↗