Tag
This paper investigates whether LLMs can infer individual domain knowledge from long-term Slack logs, comparing seven models and finding Gemini 2.5 Flash achieves the lowest error, highlighting feasibility and limits of automated expertise mapping.