- Title
- Smart knowledge engineering for cognitive systems
- Creator
- Silva de Oliveira, Caterine
- Relation
- University of Newcastle Research Higher Degree Thesis
- Resource Type
- thesis
- Date
- 2021
- Description
- Research Doctorate - Doctor of Philosophy (PhD)
- Description
- Cognition in computer sciences refers to the ability of a system to learn at scale, reason with purpose, and naturally interact with humans and other smart systems, much like humans do. Indeed, cognitive systems (CS) have emerged as an attempt to in some way mimic the capabilities of the human brain. Intelligence in cognitive computing can be considered as the capacity to learn, to apply knowledge and adapt to new situations, and to make wiser decisions as it evolves. Rather than being systematically programmed for all possible scenarios and situations, these systems learn and reason from interactions with their surroundings, through collaboration, and from experience. To enhance intelligence, as well as to introduce cognitive functions into machines, recent studies have brought humans into the CS loop, turning the system into a human–AI hybrid. To effectively integrate and manipulate hybrid knowledge for cognitive applications, suitable technologies and guidelines are required to sustain the human–AI interface so that communication can occur. Methods for gathering knowledge, formalising, learning, and reasoning about events for decision making are required to direct CS towards purposeful behaviour, and to ensure adaptability, explainability, extendability, and trustworthiness – key features of CS. However, traditional Knowledge Management (KM) and Knowledge Engineering (KE) approaches encounter new problems when dealing with cutting-edge technologies, imposing impediments for the use of traditional methods in cognitive applications. Therefore, old-style KE techniques must be extended by applying breakthroughs in emerging technology, such as new trends in supervised machine learning (ML) algorithms, knowledge representation (KR), information retrieval, learning methodologies, reasoning, etc. – all of which are called here Smart Knowledge Engineering for Cognitive Systems (SKECS). SKECS is based on methods, technologies, and procedures that bring innovations to the fields of KE, KM, and CS. The goal is to bridge the gap in the hybrid cognitive interface by the use of deep learning, experience-based knowledge representation, context-aware indexing/retrieval, active learning with a human-in-the-loop, and stream reasoning. In this work Set of Experience Knowledge Structure (SOEKS) and Decision DNA (DDNA) is extended to the visual domain and utilized for knowledge capture, representation, reuse, and evolution. The reasons for the consideration of using SOEKS with DDNA as carrier for decision making, is founded on the fact that experience must be taken into consideration for developing a cognitive system. The suggested adaptation makes these technologies suitable for use in CS and is a major contribution of this research. In addition, a range of approaches are examined throughout the layers of SKECS for applications in knowledge acquisition, formalization, storage/retrieval, learning, and reasoning, with the final goal of achieving knowledge augmentation (wisdom) in CS. Through a case study in the vision domain, a human–AI Cognitive Vision System Platform for Hazard Control (CVP-HC) is proposed. The CVP-HC addresses the current limitations of computer vision systems by bridging the gap between top-down and bottom-up approaches and enabling cognitive functions. The result is a scalable yet adaptable system capable of working in a variety of video analysis scenarios with transparency and confidence, while meeting specific industry safety requirements by modifying its behaviour accordingly. The platform serves as an environment in which investigations and evaluations of the features of SKECS can be conducted. Practical implementations based on CVP-HC shows that methods, techniques and procedures comprising the SKECS are suitable for advancing systems towards augmented cognition; they also have potential in the field of Workplace Health and Safety (WHS).
- Subject
- cognitive systems; knowledge engineering; knowledge management; SOEKS; DDNA; human-AI; augmented intelligence
- Identifier
- http://hdl.handle.net/1959.13/1430573
- Identifier
- uon:38858
- Rights
- Copyright 2021 Caterine Silva de Oliveira
- Language
- eng
- Full Text
- Hits: 924
- Visitors: 1387
- Downloads: 543
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | ATTACHMENT01 | Thesis | 3 MB | Adobe Acrobat PDF | View Details Download | ||
| View Details Download | ATTACHMENT02 | Abstract | 465 KB | Adobe Acrobat PDF | View Details Download |