Efficient Development of Parallel NLP Applications

Parallel programming is becoming increasingly popular. Computers have increasingly many cores (processors). Also, large computer-clusters are becoming available. But there is still no good programming framework for these architectures, and thus no simple and unified way for NLP applications to take advantage of the potential speed up. In this paper, we develop a broadly applicable parallel programming method to NLP problems. Our work is in distinct contrast to the tradition of designing (often ingenious) ways to speed up a single algorithm at a time. Specifically, we show how the problems which can be expressed in LBJ framework [13] take advantage of parallelization. We use Charm++ [7] to demonstrate the speed up of NLP applications.