LLMs will not replace Automated keyword extraction Kimberley Huizing | Lambda Days

[CodeSync]

LLMs will not replace Automated keyword extraction Kimberley Huizing | Lambda Days 2025

LLMs will not replace Automated keyword extraction Kimberley Huizing | Lambda Days 2025

This talk was recorded at Lambda Days in June 2025. If you’re curious about our upcoming event, check https://lambdadays.org

LLMs will not replace Automated keyword extraction

I will compare automated keyword extraction, an NLP technique, to LLM’s ability to summarise data, discuss the differences and show there are still use cases where automated keyword extraction is more suitable. I will also discuss my experiences implementing the YAKE algorithm in Clojure.

Outline of the talk:

• Explain automated keyword extraction,
• Overview of various algorithms,
• explain YAKE algorithm,
• compare YAKE vs LLM on matters like usage, resource intensiveness, data requirements, etc
• Implementing YAKE in Clojure,
• Discuss why Clojure & my experiences on this project,
• Show YAKE output for the same topics from a Reddit scraped data set and a LinkedIn scraped data set,
• Ask ChatGPT to tell me what each platform is saying on these topics,
• Conclusion: LLMs feel like talking to a human and more anecdotal. YAKE results are precise & suitable for data analysis
• Use cases for automated keyword extraction

Comments welcome! View the code-sync tag for more Code Sync talks!