STO-Activities: (no title)
Activity title:
Semantic Representation to Enhance Exploitation of Military Lessons Learned
Activity Reference:
SAS-IST-179
Panel:
SAS
Security Classification:
NATO UNCLASSIFIED
Status:
Active
Activity type:
RTG
Start date:
2023-01-31T00:00:00Z
Actual End date:
2026-01-31T00:00:00Z
Keywords:
AI, Deep Learning, Knowledge Management, Lessons Learned, NLP, Semantic Representation, SocialLinguistics
Background:
In 2018, the NATO Joint Analysis and Lessons Learned Centre (JALLC) conducted two exploratory research studies looking into the potential of modern data science techniques to support Lessons Learned (LL) analysis tasks. LL are typically documented in unstructured or semi-structured text and the hypothesis was that Natural Language Processing (NLP) techniques may help analysts to detect trends, improve the quality and utility of information held in LL repositories, and identify undocumented lessons in formal and informal communications. A key finding of these studies was that commercial-off-the-shelf NLP performed very badly when applied to NATO texts because the systems lacked the necessary semantic representation to understand the LL in a NATO context. It was recommended that a NATO LL ontology should be developed in order to enable the LL community to benefit from modern data science techniques. As JALLC has continued to explore ways to innovate in LL, under the remit of the NATO LL Capability Improvement Roadmap 2020-2025, a need for a wider range of scientific perspectives and expertise to tackle this difficult problem has been identified. Specifically, more understanding about how semantic representation can be used to enhance LL exploitation is needed in order to prepare future LL systems for the demands of AI-driven modern warfare.
Objectives:
Enhance understanding of how to achieve the rapid, reliable, and robust automation of key LL activities, such as capture, dissemination, causal analysis, classification, summarization, search, clustering, and trend analysis of military LL, in an explainable way.
The focus areas will be on understanding how to:
• Enhance semantic representations using AI, and vice versa.
• Exploit semantic representations in military LL systems.
• Integrate socio-linguistic understanding of military LL language into semantic representations.
The objectives are to:
Objective 1: Develop use cases for semantic representations in military LL systems.
Objective 2: Develop position papers that explore key concepts, e.g., modeling and reasoning over LL, LL as a KM concept, Definition of LL “nuggets”, human expert labeling/input, notions of distance/surprise/novelty, integration with operational systems, utilization of a broad spectrum of data to capture knowledge.
Objective 3: Develop publicly releasable problem statement(s) and challenge dataset(s) with evaluation metrics, that can be used by NATO or nations to launch their own research or hackathon/innovation challenges.
Objective 4: Develop (partial) semantic representations and demonstrate them applied to (elements of) the use cases.
Topics:
- The scientific topics are investigating possible ways to narrow the gap between human understanding and machine representation of information and knowledge.
- All topics aim at increasing the automation of tasks along the entire lifecycle of knowledge management.
- Exploration of semantic representations and embeddings, for guaranteeing a balance between scalability and expressivity.
- The recent advances in the projection of semantics on well-chosen mathematical spaces extend the expressivity for representing information and knowledge.
- Exploring possible combinations of semantic representations that are able to make use of the advantages of approaches from different domains of AI research.
- A thematical focus will be put on the identification and representation of causal relations within knowledge bases.
- Design of a dataset and evaluation metrics for training, testing and validating AI models.
- Use of Natural Language models to stimulate adaptation of semantic representations.
- The effects of socio-linguistic contexts resulting from different military domains
- Definition/adoption of a “notion of distance” within the context of semantic representation of KM, with the purpose of identifying novel knowledge to be integrated in LL.
Contact:
Open2Partners:
Title:
Created at 25/10/2022 18:00 by System Account
Last modified at 16/05/2024 20:00 by System Account
Go back to list
Home(NATO STO)