SBIR-STTR Award

Applied Technical Systems, Inc. (ATS) proposes to leverage its expertise with maintenance and logistics support of Naval systems and its experience in developing data-rich software solutions to research, design, and develop a data integration and analysis system that is capable of fusing unstructured (textual) and structured (machine-readable) data sources to provide a quantitative assessment (score) of the effectiveness of Naval maintenance and sustainment processes. Assessing whether existing systems and procedures are meeting their goals requires bringing together data from across multiple, disconnected silos. Much of the vital data that identifies key performance indicators, thresholds for acceptable quality or cost, or desirable outcomes are only available in unstructured text documents. We propose a system that allows users with no prior experience in Natural Language Processing (NLP) or Information Extraction (IE) practices to interact with an automated system that learns over time how to extract the data they require. This data, in turn, feeds a scoring capability that allows users to construct and monitor quantitative measures of the effectiveness of existing or planned sustainment practices. Such data can then been used to provide meaningful, objective information to support process improvement initiatives.

Benefit:
Our proposed solution provides a system that allows users to extract requirements, performance criteria, and other desirable quantities from unstructured text sources. For many organizations, key requirements are often only recorded in such documents but never explicitly recorded as quantifiable metrics. We envision that such a system that is capable of

Keywords:
key performance indicators, key performance indicators, text mining, Information Visualization, Natural Language Processing, Data Integration, text extraction, Machine Learning, Information Extraction

The purpose of our Phase II effort is to help surface actionable and quantifiable data from the text-heavy documents and files currently used by program management office staff, to expose that data in a machine-readable format that maintains a persistent anchor to the text from which it came, and to allow that staff to make faster and better informed decisions based on that data in light of their objectives. We propose in Phase II to implement a web-based collaborative software application that consolidates and exposes this information, using the designs developed and evaluated during our Phase I initial research.

Benefit:
Our proposed solution provides a system that allows users to extract requirements, performance criteria, and other desirable quantities from unstructured text sources. For many organizations, key requirements are often only recorded in such documents but never explicitly recorded as quantifiable metrics. We envision that such a system that is capable of

Keywords:
key performance indicators, Machine Learning, Natural Language Processing, text mining, Information Extraction, requirements management, Information Visualization, text extraction