ZeroVerify: An AI Framework for Automated Admission Form Verification using Zero-Shot Algorithm

Abstract

As it is known to all that admission process is one of the most critical academic activities in any university or higher education institutions requiring efficient handling of large volumes of application forms, supporting documents. Higher education institutions handle thousands of forms each year, resulting in manual verification bottlenecks, delays, inaccuracies, and fraud risks. This paper introduces ZeroVerify, a unique AI system that automates form verification with a zero-shot learning method, eliminating the requirement for task-specific training data. The architecture includes optical character recognition (OCR), large language models (LLMs) for zero-shot field extraction, cross-document reconciliation, and a declarative rules engine for eligibility verification. The proposed model is implemented in python and uses Hugging face transformers and lang chain to process documents such as transcripts, admission forms, IDs, and category related certificates.

On a dataset of 1,000 online submitted forms, ZeroVerify obtains a 93% F1-score for extraction, 94% verification accuracy, and 91% AUC for fraud detection, surpassing rule-based (70% F1) and fine-tuned baselines (85% F1) while decreasing latency to 3.8 seconds per form. This effort advances scalable, explainable artificial intelligence for educational administration.