Log data are typically stored in databases as schema-based entries in a structured format. Conventionally, log exploration requires an understanding of the fields, schema, and query parameters of the database. This disclosure describes techniques that use tabular large language models (LLMs) to process, mine, and make log data amenable to natural language queries. A relatively unsophisticated user with no database skills can query log files using natural language search. The LLMs can be fine-tuned using prompt engineering and causation information. The conventional, tedious mining of logs across multiple systems using database queries is replaced by a simple natural language interface that provides the ability to determine meaningful relationships and context across events captured within the logs. Natural language queries can enable help desks to do a basic level of troubleshooting, saving time for administrators. As more information gets added, querying and analytics of logs are simplified, with a resultant improvement in the speed and quality of troubleshooting.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Bhaskar S, Hari, "Log Exploration and Analytics Using Large Language Models", Technical Disclosure Commons, (July 27, 2023)