Noeon Research
LinkedIn
Noeon Research
Our Research
To develop an architecture that combines explicit knowledge representation in discrete structures with ML-assisted algorithms to manipulate those structures we are conducting research in the following directions:
Knowledge Representation
(01)
Proper Knowledge Representation allows us to store knowledge in graph-like structures to efficiently extract, remove or reason on it.
research_photo
Directions: Knowledge Graphs, Ontologies, Semantic Networks, Syntactic Representations, Context Hierarchy
Graphs and Reasoning
(02)
We consider reasoning to be a graph manipulation process. Thus, graph-theoretic research is crucial for fast and reliable reasoning.
research_photo
Directions: Graph Algorithms, Pattern Recognition and Substitution, Fast Retrieval, Representability, Graph Theoretical Properties, Black-Boxing and Abstraction
Agentic Behaviour
(03)
This line of research is needed to make our system agentic enough to be helpful but not agentic enough to be harmful. Reliable and aligned goal-pursuing behaviour is our priority, and we want a sane conceptualisation of it, contrary to, for example, the reward mechanism.
research_photo
Directions: Agency Theory, Goals and Intentions, Planning, World Modeling, Modal Logic
Elaboration
(04)
Elaboration of user requests is an equally important part of problem-solving. Understanding the specification of the problem, spotting and dealing with the lack of knowledge, and providing explanations require a proper interface to bridge the user's language and internal representation.
research_photo
Directions: NLP, Structurisation, Grounding, Entity Alignment, Knowledge-Guided Dialogue Systems, Pragmatic Communication
Domain-Specific Knowledge Population
(05)
The architecture is focused on abstract reasoning. It means that the system's abilities are not based on domain-specific knowledge, and to be useful the system must be augmented with a proper ontology.
research_photo
Directions: Commonsense Datasets Refinement, QoL Tools for Knowledge Management, Knowledge Distillation from LLMs
Blog
Machine Learning Assisted Graph Algorithms
What is the fastest way to do subgraph matching for knowledge extraction and meaning grounding?
Dec 06, 2023
blog-post-image
Join Our Team
We are looking for specialists for full-time employment
Read more
nodes photo
© 2023 Noeon Research. All rights reserved.
Midtown Tower 18F, 9-7-1 Akasaka, Minato-ku, Tokyo, Japan
(01)
Show email
LinkedIn