NLPBench: A Tool for Studying the Architectural Characteristics of Natural Langauge and Speech Applications

The study of architectures for natural language and speech applications has been necessitated by the large scale proliferation of such applications into desktop and embedded computers. This paper proposes NLPBench a benchmark for analyzing statistical natural language applications. This study validates the need for such a benchmark by demonstrating the lack of such applications in the currently existing benchmarks. A comparison between architectural characteristics of the applications in NLPBench is done with currently existing benchmarks to further prove the need for such a benchmark