Kitchener, ON, Canada, Feb. 11, 2026 (GLOBE NEWSWIRE) -- ApplyBoard, a leading education technology company, announced today that its research team has been awarded the Best Paper Award at the IEEE 15th International Conference on Pattern Recognition Systems (ICPRS 2025), held in London, UK. The award-winning paper, titled "Embedding Confidence to Enhance Trust in AI Document Entity Extraction," addresses a critical challenge in artificial intelligence: ensuring the accuracy and reliability of AI-generated outputs in high-stakes applications.
The research team, Matthew MacDonald, Sina Khosravi, Arash Ramin, and Sina Meraji, developed an innovative verification system that dramatically improves the trustworthiness of large language models (LLMs) when processing student transcripts, resumes, and other unstructured documents. Their solution tackles the "silent failure" problem, where AI systems generate incorrect information without indicating uncertainty.
The Innovation
While LLMs have transformed how organizations extract data from documents, they can hallucinate incorrect information, such as misreading a GPA or incorrectly transcribing course grades, without flagging potential errors. ApplyBoard's breakthrough technology acts as a quality assurance layer, analyzing the mathematical embeddings of AI outputs to assign confidence scores to every piece of extracted data. In rigorous testing, the system achieved a 98% accuracy rate (F1-score) in distinguishing between correct extractions and errors.
"This recognition from the IEEE validates our commitment to making AI not just powerful, but trustworthy," said Sina Meraji, Vice President Product Development, ApplyBoard. "In education technology, accuracy is essential, it's essential. A single data error can affect a student's admission decision and their future."
Real-World Impact
The technology enables ApplyBoard to implement a "traffic light" system that prioritizes applications for human review based on confidence scores. High-confidence extractions are flagged as low-priority for review, while uncertain data fields are immediately escalated for closer examination. This approach optimizes reviewer time by directing human attention where it's needed most, while maintaining human oversight for all applications.
The impact is substantial: the system will process hundreds of thousands of student applications annually across the ApplyBoard platform. By detecting hallucinations early in the workflow, ApplyBoard expects to reduce turnaround times from days to as little as one day, accelerating the time students receive admission offers from partner institutions.
Implementation is scheduled to begin in Q1 2026, with integration into ApplyBoard's production pipelines.
Industry Leadership in Responsible AI
The IEEE International Conference on Pattern Recognition Systems brings together researchers and practitioners from computer science, engineering, mathematics, and machine learning to advance pattern recognition technologies. Papers presented at the conference undergo rigorous double-blind peer review by an international panel of experts before publication in IEEE Xplore.
According to the IEEE conference peer reviewers, the research represents 'a novel, practical, and highly efficient method for confidence estimation' that 'significantly outperforms baselines.' The review panel praised the work for addressing 'a well-known limitation of LLMs: their inability to provide calibrated confidence scores,' noting that the approach 'requires only a single inference pass, making it highly suitable for real-world deployment.'
About ApplyBoard
ApplyBoard empowers students around the world to access the best education by simplifying the study abroad search, application, and acceptance process to more than 1,500 institutions across Canada, the United States, the United Kingdom, Australia, Ireland and Germany. Headquartered in Kitchener, Ontario, Canada, ApplyBoard has helped over 1.3 million students from more than 150 countries along their educational journeys since 2015. To learn more, visit www.applyboard.com.
Raveena Desai ApplyBoard 226-220-5319 raveena.desai@applyboard.com