With the rise of ChatCPT as the biggest Large Language Model (LLM) gaining extreme popularity, the importance of explainable Artificial Intelligence (AI) rose as well. I have been fascinated by these developments and explored (and enjoyed) the possibilities with ChatGPT myself too. The outcomes are impressive, but the possibilities to trace back the creation of these outcomes are concerning. A promising alternative for LLMs is the use of knowledge graphs to create natural language. Knowledge graphs are networks that store information based on a clear set of rules. They store bits of information in nodes and relations are represented by their edges. This enables true transparency on where the output is based on and flexibility regarding correcting the stored information. Sony CSL has given me the opportunity to investigate this process. During my internship I will write my master thesis about replicating previous work that created narrative structures based on knowledge graphs. I will evaluate the process and try to improve it.
