Research Publications

Permanent URI for this communityhttps://rda.sliit.lk/handle/123456789/4194

This main community comprises five sub-communities, each representing the academic contribution made by SLIIT-affiliated personnel.

Browse

Filters

20262

Advanced Search

Filter by

Settings

Subject: search.filters.subject.bias detection

Search Results

Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemEmbargo
    LegalVision: A Knowledge-Driven AI Framework for Legal Understanding and Trust Assessment
    (Institute of Electrical and Electronics Engineers, 2026-06-12) Sharan K; Wicramasinghe D.A.T.N.; Maxwell L.Y; Sivanuja S; Kuruppu, D.S; Dissanayake, A
    Legal documents are often lengthy, complex, and written in highly technical language, making them difficult for citizens and even legal professionals to interpret efficiently. This creates a need for an intelligent legal support system that can improve accessibility, transparency, and trust in document understanding. This study proposes LegalVision, a knowledgedriven AI framework that integrates multi-perspective legal summarization and visualization, explainable legal reasoning, clause-level bias and risk evaluation, and a dynamic legal knowledge graph for property law documents. This research is conducted within the Sri Lankan legal context using a dataset collected from real Sri Lankan legal documents, including property-related deeds and agreements. The framework processes legal texts through clause segmentation, entity and relation extraction, perspective-based summary generation, infographic visualization, risk classification, and graphsupported reasoning, while preserving links to the original clauses for traceability within a single platform. Therefore, this research contributes a unified and explainable legal AI framework that supports both legal professionals and nonexpert users in understanding property law documents more accurately and efficiently.
  • Thumbnail Image
    PublicationOpen Access
    EuqAud: Detecting Gender Bias in Audio Datasets Using Polynomial Regression-Based Metric
    (Institute of Electrical and Electronics Engineers Inc., 2026) Jayawardena, S; Haddela, P.S; Shyamalee, T; Ekanayake, A; Mudalige, T; Dhanawardhana, I
    With the growing adoption of audio based AI systems in high-stakes domains such as healthcare, law enforcement, and social media, ensuring fairness particularly regarding gender bias has become critically important. While prior work on fairness has predominantly focused on disparities in model performance, bias inherent in training datasets remains underexplored. To address this gap, we propose EuqAud, a novel, pre-trained and traceable fairness metric that quantifies gender bias in audio datasets using raw acoustic features such as pitch, energy, amplitude, and voice activity. Unlike methods dependent on demographic labels such as race, age or language, EuqAud is designed to be demographic and language agnostic, enhancing its applicability across diverse contexts. The score is computed using an equation derived from polynomial regression with L2 regularization (Ridge regression), yielding robust and generalizable outputs. It spans a range from −10 to 10, where 0 denotes neutral, positive scores indicate male dominant bias, and negative scores reflect female dominant bias. For clarity, bias severity is categorized into three tiers: Neutral (EuqAud < 2), Moderate Bias (2 ≤ EuqAud ≤ 6), and Strong Bias (EuqAud > 6). Evaluation across multiple datasets demonstrates high predictive performance, with R2 values between 0.95 and 0.99. By focusing on dataset level bias rather than model outcomes, EuqAud offers a scalable and rigorous solution for advancing fairness in audio-based AI systems.