Finding Concerts using Large Language Models : The Stockholm Concert Database as Case Study for ’Touringbot’
(2024) 6th Digital History in Sweden Conference- Abstract
- This presentation presents a method for using Large Language Models (LLMs) to structure historical data about Swedish musical history. The research builds on the existing, human-created Stockholm Concert Database of concerts occurring in Stockholm between 1848-1908. Using this as a starting point, it details an experiment in automated extraction of information about concerts from historical Swedish newspapers, explaining the digital humanities methods that were used to access data from the National Library of Sweden, and the new methods being used to process this data using LLMs, including prompting strategies and comparison of the accuracy of different LLMs. The presentation then demonstrates how the accuracy of this new approach is... (More)
- This presentation presents a method for using Large Language Models (LLMs) to structure historical data about Swedish musical history. The research builds on the existing, human-created Stockholm Concert Database of concerts occurring in Stockholm between 1848-1908. Using this as a starting point, it details an experiment in automated extraction of information about concerts from historical Swedish newspapers, explaining the digital humanities methods that were used to access data from the National Library of Sweden, and the new methods being used to process this data using LLMs, including prompting strategies and comparison of the accuracy of different LLMs. The presentation then demonstrates how the accuracy of this new approach is measured by comparing it to existing results from the Stockholm Concert Database.
Touringbot underscores the potential for a major leap in the scale and cost of historical datasets: While the original database was resource-limited to a “sliced-history” approach to sampling only one year per decade, this automated method is able to fills in the gaps. These preliminary results are intended to demonstrate the feasibility and promise of this new method for applications in the study of history, as well as to highlight remaining hurdles to be overcome. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/d3115d61-eec8-4f94-aed5-4e0c885f276c
- author
- Farnsworth, Brandon
LU
- organization
- publishing date
- 2024
- type
- Contribution to conference
- publication status
- unpublished
- subject
- conference name
- 6th Digital History in Sweden Conference
- conference location
- Växjö, Sweden
- conference dates
- 2024-11-07 - 2024-11-09
- language
- English
- LU publication?
- yes
- id
- d3115d61-eec8-4f94-aed5-4e0c885f276c
- date added to LUP
- 2025-03-03 11:12:32
- date last changed
- 2025-04-04 13:53:02
@misc{d3115d61-eec8-4f94-aed5-4e0c885f276c, abstract = {{This presentation presents a method for using Large Language Models (LLMs) to structure historical data about Swedish musical history. The research builds on the existing, human-created Stockholm Concert Database of concerts occurring in Stockholm between 1848-1908. Using this as a starting point, it details an experiment in automated extraction of information about concerts from historical Swedish newspapers, explaining the digital humanities methods that were used to access data from the National Library of Sweden, and the new methods being used to process this data using LLMs, including prompting strategies and comparison of the accuracy of different LLMs. The presentation then demonstrates how the accuracy of this new approach is measured by comparing it to existing results from the Stockholm Concert Database. <br/>Touringbot underscores the potential for a major leap in the scale and cost of historical datasets: While the original database was resource-limited to a “sliced-history” approach to sampling only one year per decade, this automated method is able to fills in the gaps. These preliminary results are intended to demonstrate the feasibility and promise of this new method for applications in the study of history, as well as to highlight remaining hurdles to be overcome.}}, author = {{Farnsworth, Brandon}}, language = {{eng}}, title = {{Finding Concerts using Large Language Models : The Stockholm Concert Database as Case Study for ’Touringbot’}}, url = {{https://lup.lub.lu.se/search/files/209565507/Digital_History_in_Sweden_Conference_Presentaiton.docx}}, year = {{2024}}, }