Analyzing and Assessing Contracts for Embedded Risk
Systems Engineering and Systems Management Transformation
Report Number: SERC-2020-TR-014
Publication Date: 2020-09-26
Project: Analyzing and Assessing Contracts for Embedded Risk
Dr. Carlo Lipizzi
This Technical Report for WRT-1023, Analyzing and Assessing Contracts for Embedded Risk, provides a summary of research activities and key findings from 27 September 2019 to 26 September 2020.
Using Natural Language Processing (NLP) via Machine Learning (ML) techniques, this research project worked to provide a data-driven computational model and developed a working prototype to support and streamline the relevant parts of the DoD contracting processes from beginning to end while.
The research effort showcased the application of data analytics in order to understand the assessment processes undertaken by a Contracting Officer (CO). Currently, the prototype boasts the Federal Acquisition Regulation (FAR), Defense Federal Acquisition Regulation Supplement (DFARS), and several other guidelines, as well as course materials to help establish how current contract types are identified. However, a separate objective was to also map the current process against best practices for the U.S. Department of Defense (DoD) contracting process from beginning to end. This would include reviews and interviews with subject matter experts (SMEs).
There have been many attempts to understand the root causes of inefficiencies in the DoD contracting process. The goal of this research effort was to apply data analytics to understand the assessment processes undertaken by a contracting officer (CO). The intent was to bring significant efficiencies to these assessment processes and to develop a prototype tool covering relevant parts of the DoD contracting process from beginning to end.
Applying data analytics to understand the assessment process undertaken COs can lead to understanding and alleviating root causes of inefficiencies in the DoD contracting process. The research leveraged experience and knowledge from a concurrent Research Task, “Meshing Capability and Threat-based Science and Technology (S&T) Resource Allocation,” to aid in developing a working system/prototype by extracting metrics from texts to evaluate key elements within the context of the text. The research is be based on grounded theory, allowing theories to emerge from collected data, and utilizes a combination of interviews, existing planning and budgeting documents, and other relevant artifacts to support a text-driven approach to model a system to make decisions.
For this development of the prototype, the research team:
- Leveraged current literature to create a logical framework to classify requests based on a given contract type
- Created a computational model for the logical framework
- Created a visualization system to present the results in a digestible manner
- Delivered results with an agile approach, developing prototypes/proofs of concepts with increasing capabilities.
The prototype is based on open source components that were integrated via research-grade algorithms and methods developed by the team for this task. The prototype, in this first year of research, proved the validity of the approach and covered the basic functionalities. However, because of its nature of being a prototype, it does have limited robustness, interactivity, productivity, and reusability. Nevertheless, the prototype proved to sufficiently stratify the appropriate contract structure.
In addition, the team was able to start developing tasks that were initially intended in further phases/years of the project. One being a web-based user interface (UI) that allows documents to be uploaded (separately or together) to determine which contract type is best fitted for the requirements. Further refinement can be achieved through greater incorporation of expert knowledge. Through interviews with stakeholders, additional documents provided by the Sponsor, and scraped open source material, the team developed key elements for the contract analysis. Nevertheless, further inclusion of user feedback, SME, and growth to the corpus will improve the prototype to be more robust.