TCSE logo 
 Sigsoft logo
Sustainability badge
Tue 29 Apr 2025 14:00 - 14:45 at 107 - Invited Talks

Data entry errors pose a significant challenge in workflows in- volving manual input of information from unstructured sources into structured formats. These errors, often arising from fatigue or oversight, lead to costly corrections and operational inefficiencies. Existing approaches, such as input validation and independent re- views, while valuable, are resource-intensive and fail to address all error types effectively. This paper introduces DocuBot, a system powered by large language models (LLMs) designed to detect data entry errors early in the process. By comparing natural language inputs with their structured representations, DocuBot identifies inconsistencies, flags potential errors, and, where possible, suggests corrections. When validation is inconclusive, it prioritizes entries for human review. This early detection reduces the reliance on traditional re- view methods, thereby lowering compliance costs and enhancing data integrity. We evaluate DocuBot through a series of experiments, demon- strating its high precision in error detection and its capacity to streamline workflows that require human oversight. While not yet suitable for full automation, DocuBot represents a promising step toward minimizing the financial and operational impacts of data entry errors.

Martin Schäf is a senior applied scientist at AWS. Before joining AWS, he worked at SRI International. Martin did his PostDoc at the United Nations University in Macau. He received his PhD from University of Freiburg in 2011, and his MS degree in computer science from Saarland University in 2006. His research interests include static analysis, software verification, fault localization, and testing.

Tue 29 Apr

Displayed time zone: Eastern Time (US & Canada) change

14:00 - 15:30
Invited TalksSTATIC at 107
14:00
45m
Talk
Automatic Detection of Data Entry Errors
STATIC
I: Martin Schäf Amazon Web Services
14:45
45m
Talk
The Pursuit of Soundness in Android Static Analysis
STATIC
I: Jordan Samhi University of Luxembourg, Luxembourg
:
:
:
: