Research Foci

Below are the main research topics for the KASTLE lab. Current, past, and possible independent studies or research project opportunities are listed below each topic. We are currently welcoming Independent Studies and Master’s Theses pertaining to the below topics listed as “OPEN”.

Helpful acronyms:

KG: Knowledge graph
LLM: Large language model
ACE: Attempto Controlled English

Human-Centric Knowledge Engineering

Generally pertaining to the methodologies for developing of knowledge graphs, data integration, and ontology alignment, using human action.

Knowledge Engineering through Prompt Engineering (OPEN) – This research will involve mining LLMs for knowledge through prompts. For example, co-reference resolution or ontology alignment.
Using LLMs for KG Construction (OPEN) – This research will involve the construction of a KG schema through ODP-based methods.
Topics in Modular Ontology Design (OPEN) – This research might be the development of new ontology design patterns.

Automated Knowledge Engineering

Generally pertaining to the development of tools, frameworks, and methods for automatically constructing ontologies, as well as conducting data integration, semantic harmonization, schema alignment, and visualization.

Exploring Lexico-syntactic Patterns (OPEN) – This research avenue pertains to understanding how lexico-syntactic patterns might empower LLMs to use graph-based data.
Dynamically Creating KG Views from Module Metadata (Antrea Christou)

Natural Language Processing and Knowledge Graphs

Generally pertaining to the extraction of formal knowledge from unstructured natural language, which is subsequently used to inform downstream tasks.

KG-powered Chat Bot (Anmol Saini)
Connecting Multiple Latent Spaces through Fixed-Points (OPEN) – This research avenue will seek to unify latent spaces generated over natural language and KGs.
Fine-tuning an LLM to output ACE (OPEN)

Neurosymbolic Artificial Intelligence

Generally pertaining to the combination of symbolic knowledge representation and reasoning with the power of neural learning systems. These topics tend to bridge between many categories.

Investigation of Hidden Layer Node Activations in NL-KG Concurrent Systems (OPEN)
Understanding the Impact of Graph Structure on KG Embeddings (Brandon Dave)
Semantically Composable Glyphs for Representing AI Systems (Alexis Ellis)

Open Science

Generally pertaining to methods where knowledge graphs and semantic technologies are used to promote open science, improve reproducibility and replicability, discovery of long-tail data.

Meta-epistemology: A Model for Understanding Knowledge Gaps (OPEN)

Effective Pedagogy

Generally pertaining to the methods for effective pedagogy of computer science, data science, knowledge graphs, Semantic Web, and knowledge engineering. Generally, this involves the following resources.

There is only a single topic, but it is quite variable, where contributing some set of articles based on interest could constitute an independent study.

Building Curriculum for Knowledge Engineering (OPEN)