NBLucene: Flexible and Efficient Open Source Search Engine

The most popular open source projects for text searching have been designed to support many features. These projects are well-written in Java for cross-platform using. But when conducting research, the execution efficiency of program should be more essential, which is a problem for applications written in Java. It is also difficult for Java to use parallel mechanisms in the modern computer system like SIMD and GPUs. To this end, we expand an open source text searching project written in C++ for research purpose. Our approach is to define a flexible and efficient search engine architecture which consists of extensible application programming interfaces. We aim to provide a flexible architecture to enable researchers to readily implement and modify search engine algorithms and strategies. Moreover, we integrate one generic mathematical encoding library which can be used to compress inverted index. We also implement an integral framework for result summarization, including snippet generation and cache strategies. Experiment results show that the new architecture makes a significant improvement versus original work.