At the start of this journey, I said I would conduct my investigations in 9 Phases:
- AI support for the Memento collection’s index entries.
- AI support for the Memento collection’s combined Index entries and file titles.
- AI support for the Memento collection’s index entries, file titles and textual items.
- AI support for PAWDOC’s index entries.
- AI support for PAWDOC’s combined index entries and file titles.
- AI support for PAWDOC’s combined index entries, file titles, and some or all of the born digital items.
- AI support for a subset of PAWDOC’s scanned items.
- AI support for a combination of index entries, file titles, some born digital material and some scanned items.
- AI support for the whole of PAWDOC.
I’ve completed the first three Phases and reported on them in previous posts. However, given what I have learned up to now, I’ve decided to change my plan in the following ways:
Combine Phases 4 and 5: It’s a feasible and better approach to obtain AI support for a combination of Index entries and associated File names.
Combine Phases 6, 7, and 8: I’ve discovered that AI can deal with scanned documents in much the same way as born digital documents, so they can both be investigated together. Furthermore, given the limits on context windows, and file numbers and sizes, it’s not really practical to try to obtain AI support for combinations of the whole index and large numbers of content files. Instead, I shall investigate the approach of applying AI to a subset of content files and their associated index entries.
Change Phase 9 – AI support for the whole of PAWDOC will have been explored in the new Phases outlined above. Instead, the final Phase will assess all my findings about how AI can support personal archives
Add another Phase: Since setting out the Phases back in March this year, I’ve discovered that the FileMaker application I use to manage the PAWDOC Index, has been incorporating progressive amounts of AI support since the version I currently have which came out in 2019. Furthermore, FileMaker is currently offering a 45-day free trial of its latest version – FileMaker Pro 2025. I intend to take up this offer once I have completed the revised PAWDOC Phases. This will include watching various training videos about FileMaker’s AI’s capabilities in addition to performing tests on the PAWDOC Index and Files. I anticipate that the way FileMaker has addressed the opportunities afforded by AI, will extend my AI knowledge as well as providing an insight into how any Database/Indexing/Repository software might try to exploit AI.
In summary the Phases I now intend to follow are as follows:
1-3: Complete.
4. AI support for PAWDOC’s combined index entries and file titles
5. AI support for selected subsets of PAWDOC born digital/scanned document and their associated index entries.
6. AI Support provided by FileMaker for the PAWDOC Index.
7. Assessing all my findings about how AI can support personal archives.