Used ChatGPT4 to rewrite what I wrote at the bottom:
* 2015: Utilized ACT-DL (Digital Classification for Librarians, discontinued in 2022) and extensive Excel analysis, focusing on probability percentages for prioritization. Analyzed personal writings from 1989-2015, stored online. Dataset: DOI: 10.17632/92825j2d2r.1 – used ~9,500 samples out of 36,000.
2021: Employed WordStat 8 for Latent Dirichlet Analysis (LDA) with 1-99 “bag of words” topic groupings. Combined groupings to identify “master topics.” Concurrently, used LibraryThing to catalog books read/owned and representative books from Wikipedia searches (2013-2019). Retrieved DDC numbers and additional data from OCLC FAST subject headings. However, lacked computational power to analyze this extra information.
2023: New data source – ChatGPT Conversation Log Export (Dec 2, 2022 – May 5, 2023). Converted logs to PDF, then to text, and performed LDA analysis online. Used ChatGPT4 to derive likely topics and equivalent Dewey Decimal numbers/names. Analyzed additional subjects from LibraryThing and WordStat LDA (2021) for representative DDCs. Excel served as the intermediary for all data, generated through AI or a mix of AI/manual curation (e.g., book classification). This summary presents a compact history of the results.