Building a Research Assistant Management Platform utilizing natural language processing

As the largest navy in the world, the United States Navy must quickly retrieve and communicate information across the globe in order to protect the country and its allies. As the amount of information that analysts must sort through grows exponentially, this creates a more difficult environment for extracting useful and relevant data. This problem of information complexity in research is present in other disciplines, creating a demand for a system to increase both effectiveness and efficiency of research. In this paper, we propose a system, the Research Assistant Management Platform (RAMP), to assist analysts in their duties in retrieving relevant maritime information. The platform is a realization of a proof of concept utilizing a database backend for storage of Requests for Intelligence, real-time updating topic models for document similarity and thematic analysis, and an intuitive front-end user interface with built-in work-flow operations such as saving and model visualizations. The system provides effective information retrieval and an integration of a previously fragmented multi-step process into one research platform. The prototype uses surrogate data which is contextually and structurally similar to metadata which would be seen by analysts. With this data, the system was evaluated based on user testing and the determined relevance of the displayed results from a query. The final topic model consists of fifty topics which allow analysts to improve their responses and learn from the existing corpus of RFIs. To the best of our knowledge, the implementation of an integrated research platform in this context is a novel application and although presented in the context of maritime research, this platform is generalizable to other commercial and academic uses.