Console Log Explainer: A Framework for Generating Automated Explanations using LLM

Log analysis is challenging due to the unstructured nature of $\log$ messages. Most prior methods are limited to specific $\log$ formats and require substantial labeled data. Recent natural language processing techniques, such as large language models (LLMs), show promise for automated log analysis. This study aims to develop a generalized framework for constructing LLMbased systems to explain technical console logs across applications. The paper introduces a novel framework to construct LLMbased log explanation systems and proposes a model combining Falcon-7b LLM fine-tuned using Parameter Efficient Fine-Tuning (PEFT) techniques on a curated log dataset. An original dataset with over 100 error logs spanning 10 applications was created to enable training and benchmarking. Human evaluation and inter-rater reliability quantification compared log explanations from the model and baselines. The proposed model achieved near perfect agreement to generate explanations (Kappa $=0.84$) that improved substantially over the baseline Falcon and other LLMs. It demonstrated accurate and comprehensible log error explanation capabilities. The study shows feasibility of generalized LLM-based systems with only a few data to automatically generate understandable explanations for diverse console error logs.