New Metric Based on SQuAD for Evaluating Accuracy of Enterprise Search Algorithms

While Coseer’s engineering team innovates on a regular basis, most of such technical innovations have to be closely guarded as a trade secret under our company policy. However, there are some gems we share from time to time. In 2020, we presented one such innovation at the Future of Information and Communication Conference. The corresponding paper has been published in its proceeds, Advances in Information and Communication.

Enterprise Search is a continuously evolving and important field, which is seeing a resurgence driven by artificial intelligence. Still, there is no objective, generally accepted way to compare various enterprise search systems. SQuAD is becoming popular for measuring algorithmic reading comprehension (MRC) but is ineffective for quantifying effectiveness of enterprise search in business-use situations. In this paper the Coseer team modifies the SQuAD scoring methodology to propose a scoring system for enterprise search systems that aligns with the real world expectations of users. They have used a search system based on Calibrated Quantum Mesh (CQM) to underscore the relevance of this metric.