Interactive Knowledge Graphs for Situational Awareness

OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Trusted AI and Autonomy The technology within this topic is restricted under the International Traffic in Arms Regulation (ITAR), 22 CFR Parts 120-130, which controls the export and import of defense-related material and services, including export of sensitive technical data, or the Export Administration Regulation (EAR), 15 CFR Parts 730-774, which controls dual use items. Offerors must disclose any proposed use of foreign nationals (FNs), their country(ies) of origin, the type of visa or work permit possessed, and the statement of work (SOW) tasks intended for accomplishment by the FN(s) in accordance with the Announcement. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws. OBJECTIVE: This topic seeks to research state of the art techniques for enabling a user to interact with a dynamic knowledge graph and suggest additional necessary changes to the graph in support of faster, more effective situational awareness, pattern of life analysis, threat detection, and targeting operations in time-constrained environments. DESCRIPTION: Knowledge Graphs capture information about entities and the relationships between those entities, represented as nodes and edges within a graph. Entities can be comprised of objects, events, situations, or concepts. Knowledge Graphs are typically constructed from various data sources with diverse types of data, creating a shared schema and context for formerly disparate pieces of data. As such, Knowledge Graphs provide a rich source of information, enabling capabilities like question and answering systems, information retrieval, and intelligent reasoning. Of interest to the Air Force are Knowledge Graphs that enable situational awareness, pattern of life analysis, threat detection, and targeting operations. Dynamic knowledge graphs are well-suited to these applications due to changing nature of an operational environment. Previous AFRL/RI research has shown the added value to analysts in capabilities that allow a user to interact with stored data, especially when the data was structured and stored by machine learning (ML) and artificial intelligence (AI) based capabilities. AI/ML approaches for identifying, structuring, and storing data are not 100% trusted by analysts, so there is a requirement to enable the analyst to make corrections to the data as needed and manually add additional data. Currently, these modifications are all entered by manual methods and are not utilized to further improve the overall data store and analysis capability, making it a time consuming task that will not be feasible in a time-constrained environment, such as a peer fight. Existing approaches for interacting with a knowledge graph have focused on visualization techniques and query mechanisms, but additional user interaction is required for such graphs to be useful and trusted by users in AF applications, especially when time of the essence. The goal of this topic is to research techniques that allow a user to interact with a dynamic knowledge graph by making changes and additions to the knowledge graph, and then utilize that user input to suggest additional updates to surrounding nodes/edges in the graph. This may include, but is not limited to, aspects such as: updating the graph's underlying ontology/schema, inferring additional edges between nodes, highlighting conflicting information in the graph, highlighting information gaps, and suggesting additional changes to the graph because of the user's modifications. Such a capability would significantly reduce the current time required to complete analytic tasks, such as those in support of targeting operations, as less time would be spent manually correcting the results of AI/ML data structuring, correlation, and storage. This work is to be done at the unclassified level for Phase I and II and offerors should plan to use their own datasets. PHASE I: Phase I awardee(s) will experiment with and assess feasibility of different approaches to allow for user modifications to a dynamic knowledge graph and predict additional necessary changes to the graph based on the user's input. In support of this, awardee(s) will obtain baseline performance metrics, such as (but not limited to) accuracy and graph completeness, of the different approaches to select most promising approach. Based on these results, develop an initial prototype design and document all work completed in this feasibility study. PHASE II: Awardee(s) will perform in-depth research and develop a full-scale prototype for adaptive, interactive dynamic knowledge graphs. Awardee(s) will demonstrate the effectiveness of the capability on an AF-relevant application. Awardee(s) will deliver software and support on-site testing in the customer's environment. Awardee(s) will evaluate performance of the prototype compared to baseline from Phase I. PHASE III DUAL USE APPLICATIONS: Phase III will pursue transition paths for military and commercial applications. Potential users may include, but are not limited to, AF intelligence analysts, law enforcement, homeland security, and financial markets. This phase will also focus on inserting and evaluating performance of the developed capability in operational environments. REFERENCES: 1. Yan, Yuchen et al. Dynamic Knowledge Graph Alignment. Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 11, 2021, pp. 10112 10119., https://ojs.aaai.org/index.php/AAAI/article/view/16585/16392; 2. Nechasky, Martin, and Stepan Stenchlak. "Interactive and iterative visual exploration of knowledge graphs based on shareable and reusable visual configurations." International Journal of Human-Computer Studies, vol. 163, 2022, pp. 102839., https://www.sciencedirect.com/science/article/abs/pii/S1570826822000105 KEYWORDS: adaptive knowledge graph; dynamic knowledge graph; interactive knowledge graph