Research

As part of S23M's partnership with AUT Colab, S23M intends to sponsor Master's and PhD students with an interest in one of the following areas of research:

• Developing assistive tooling for human cognition and thinking

  Knowledge representations that are intuitive and easily understandable for humans and easily processed by software tools, and corresponding context-aware semantic search tools.

  • PhD topic 1: Visualisation of semantic artefacts on internet connected devices

• Creating a knowledge distillery and related feedback loops

  Representations of knowledge flows at all levels of scale and between levels of scale: agents/teams/organisations/communities; techniques and tools for collaborative validation of knowledge and beliefs; techniques and tools for merging human domain knowledge with knowledge extracted from data by machine learning; principles and techniques for interdisciplinary innovation and the creation of context specific variants of bodies of knowledge.

  • PhD topic 2: Tool assisted conversion of textual domain knowledge into formal models
  • PhD topic 3: Unsupervised machine learning techniques that produce human understandable representations

PhD topic 1: Visualisation of semantic artefacts on internet connected devices

This research would explore new paradigms for the specification of visual user interfaces that can be easily mapped to the capabilities of the increasingly diverse kinds of internet connected devices. The research would relate to open source generic artefact visualisation software that S23M is developing for its open source Cell Platform technology.

Challenges:

1. Creating a visual language framework for semantic artefacts that can be used across different internet connected devices and that allows users to view the names of semantic identities in their preferred language
2. Exploiting the specific advantages of specific user interface technologies (touch screens, speech recognition, eye tracking, and further emerging technologies such as radar based hand gesture recognition)
3. Keeping the visual language framework for semantic artefacts consistent across the different user interfaces
4. Adapting semantic artefact representations and the user interaction model for navigating semantic artefacts to the limits imposed by available screen real estate
5. Integrating suitable tools for development of domain specific symbols, and providing functionality that allows human domain experts to share visual symbols and engage in collaborative symbol development and refinement
6. Developing an architecture that is designed to cope with growing variability across internet connected devices, and using the Cell Platform to achieve semantic interoperability across different implementation technologies

PhD topic 2: Tool assisted conversion of textual domain knowledge into formal models

Research into tool assisted conversion of textual domain knowledge into formal models. This research would involve the development of innovative approaches to natural language processing, new paradigms for knowledge sharing between humans and machines, and make use of the advanced semantic modelling capabilities inherent in S23M's open source Cell Platform technology.

Challenges:

1. Offering a range of different ways of to provide knowledge mining tools with semantic context information in relation to specific sets of textual input artefacts, for example in the form of semantic domains and category models articulated in Cell technology, or in the form of dictionaries or ontologies and tools to transform these into formal semantic domains and category models
2. Compiling a comprehensive set of realistic use cases, and for each use case to define appropriate heuristics for combining automated and (manual) human steps for knowledge extraction, including validation of the heuristics with domain experts from a range of disciplines
3. Developing tools for visualising semantic equivalences and differences in preferred terminologies between human domain experts, as well as tools for consolidating semantic domains and differences in terminologies
4. Developing an architecture that enables tool assisted extraction of domain knowledge from textual input artefacts that may be expressed in a variety of languages (English, Chinese, Spanish, etc.)
5. Developing automatic suggestions for the use of specific terminologies and jargons based on the social network of agents that a user is interacting with
6. Development of analytical tools to detect terminological drift and to track the adoption of new words in the context of specific disciplines and semantic domains

PhD topic 3: Unsupervised machine learning techniques that produce human understandable representations

Research and development of advanced unsupervised machine learning capabilities that lead to representations of knowledge that are human understandable, and that can easily be integrated with formal representations of the knowledge of human domain experts. The neural networks and algorithms developed as part of this work would be expressed as semantic artefacts in S23M's open source Cell Platform technology.

Challenges:

1. Providing machine learning tools with semantic context information in the form of semantic domains and category models articulated in Cell technology, and implementing machine learning tools and algorithms that are capable of using such context information in unsupervised learning mode
2. Using Cell technology features to automatically generate semantic artefacts that quantify conformance with semantic context information provided by human domain experts
3. Using graphical probabilistic models and available semantic context information to suggest new semantic identities that have been detected by machine learning algorithms
4. Developing suitable visualisation and user interaction patterns that assist human domain experts in the naming of new semantic identities that have been created by machine learning tools
5. Compiling a comprehensive set of realistic use cases, and for each use case to define appropriate heuristics for combining machine learning and (manual) human steps for knowledge creation and validation
6. Developing tools for visualising semantic equivalences and differences in models, as well as tools for consolidating model differences across viewpoints, based on appropriate explicit agreement by all human domain experts that are affected by the consolidation