query is an important and essential operator for data analysis over string collections. However, when uncertainty comes into big data, it calls for new parallel algorithms for efficient query processing on large scale uncertain strings. In this paper, we proposed a MapReduce-based parallel algorithm, called MUSK, for answering top-
queries over large scale uncertain strings. We used the probabilistic
-grams to generate key-value pairs. To improve the performance, a novel lower bound for
expected edit distance
was derived to prune strings based on a new defined function
gram mapping distance
. By integrating the bound with TA, the filtering power in the Map stage was optimized effectively to decrease the transmission cost. Comprehensive experimental results on both real-world and synthetic datasets showed that MUSK outperformed the baseline approach with speeds up to 6 times in the best case, which indicated good scalability over large datasets.