Muhammad Saad Amin

This project conducts applied research on integrating Large Language Models (LLMs) into Software Project Management (SPM) by designing and evaluating GenAI-based systems to support core SPM tasks.

Leveraged Large Language Models (LLMs) to enhance social impact by improving decision-making processes, considering user information from various perspectives.

Traditional metrics often miss structural and linguistic fidelity, so we introduce two bidirectional evaluation methodologies, Pars-Gen and Gen-Pars, for comprehensive DRS-based system assessment for semantic and structural assessments.

Semantic parsing and text generation with DRS are reversible but error-prone. We leverage this reversibility to investigate error propagation and explore error correction through innovative pipeline approaches from multi-lingual perspective.

A novel approach of augmenting formal meaning representation (DRS) through Named Entity (PER, GPE), Lexical (nouns, verbs, adjectives, adverbs), and Grammatical (tense) augmentation strategies for semantic processing. Furthermore, the utilization of data augmentation for semantic parsing and text generation for English, Italian, and Urdu.

Development of the first Semantic Resource for Urdu and its utilization for neural semantic parsing and generation tasks with the adaptation of data augmentation.

Investigated data delexicalization techniques to improve model generalization, addressing the challenge of heavily integrating lexical entities with external DRS resources, and emphasizing structural sequences over lexical knowledge.

A novel approach of using Super Senses to perform data augmentation based on the lexical categories of nouns, resulting in four flavors of augmentation. e.g., lexical substitutions through inside/outside context and with/without supersenses.

Developed a smart sensor-based data glove prototype for capturing alphabetical and numerical sign postures. Generated datasets and utilized neural networks for classifying static sign gestures.