People increasingly search online for answers to their medical questions but the rate at which medical questions are asked online significantly exceeds the capacity of qualified people to answer them. This leaves many questions unanswered or inadequately answered. Many of these questions are not unique, and reliable identification of similar questions would enable more efficient and effective question answering schema. COVID-19 has only exacerbated this problem. Almost every government agency and healthcare organization has tried to meet the informational need of users by building online FAQs, but there is no way for people to ask their question and know if it is answered on one of these pages. While many research efforts have focused on the problem of general question similarity, these approaches do not generalize well to domains that require expert knowledge to determine semantic similarity, such as the medical domain. In this paper, we show how a double fine-tuning approach of pretraining a neural network on medical question-answer pairs followed by fine-tuning on medical question-question pairs is a particularly useful intermediate task for the ultimate goal of determining medical question similarity. While other pretraining tasks yield an accuracy below 78.7% on this task, our model achieves an accuracy of 82.6% with the same number of training examples, an accuracy of 80.0% with a much smaller training set, and an accuracy of 84.5% when the full corpus of medical question-answer data is used. We also describe a currently live system that uses the trained model to match user questions to COVID-related FAQs.

The goal of this work is to estimate the accuracy of a classifier on a large unlabeled dataset based on a small labeled set and a human labeler. We seek to estimate accuracy and select instances for labeling in a loop via a continuously refined stratified sampling strategy. For stratifying data we develop a novel strategy of learning r bit hash functions to preserve similarity in accuracy values. We show that our algorithm provides better accuracy estimates than existing methods for learning distance preserving hash functions. Experiments on a wide spectrum of real datasets show that our estimates achieve between 15% and 62% relative reduction in error compared to existing approaches. We show how to perform stratified sampling on unlabeled data that is so large that in an interactive setting even a single sequential scan is impractical. We present an optimal algorithm for performing importance sampling on a static index over the data that achieves close to exact estimates while reading three orders of magnitude less data.