Developing a modular framework for implementing a semantic search engine

Abstract

Current methods of information retrieval (IR) are adequate for everyday search needs, but they are not appropriate for many military and industrial tasks. The underlying mechanism of typical search methods is based upon keyword matching, which has demonstrated very poor performance over highly technical requirements documents found within the field of acquisitions. Instead of matching keywords, IR methods that understand the meaning of the words in a query are needed to provide the necessary performance over these types of documents; this is known as semantic search. This work utilizes sound software engineering practices to specify, design, and develop a modular framework to aid in the design, testing, and development of new semantic search methods and IR techniques, in general. The development of Modular Search Engine framework is documented in its entirety, from user needs analysis to the production of a full application programming interface. By exploiting the powerful techniques of polymorphism and object-oriented programming in the Java programming language, users are able to design new IR techniques that will function seamlessly within the framework. Finally, a reference implementation is provided as a proof-of-concept to demonstrate the capabilities and usefulness of the framework design.