AI/ML - text summarization

[zoometh]

I have consistently maintained two primary DOCX files (Sites.docx and Cultures.docx) over the years, each comprising approximately 1,500 pages. These documents are primarily written in French and focus on the European Pre- and Protohistory, specifically detailing archaeological sites (e.g., Stonehenge) and archaeological cultures or periods (e.g., Neolithic). The content includes tables, cross-references (hyperlinks and anchors between Sites.docx and Cultures.docx), bibliographic reference keys, and idiosyncratic notations (e.g., EN for Early Neolithic, BIB followed by a number for a bibliographic reference). Additionally, the files are organized with titles and subtitles defining:

• temporal groupings (e.g., Neolithic covering Early, Middle, and Late Neolithic);

• spatial groupings (e.g., Early Neolithic Atlantic, Early Neolithic West Mediterranean);

• typological groupings (e.g., megalithism covering dolmen, menhir, and tumuli; weapons covering axes, spears, swords, etc.).

I am seeking extractive summarization through prompts capable of managing the intricate structure of my archives. For instance, feeding the LLM with these two documents I would like to receive summaries such as:

Resume the different chapters on funerary practices during the Early Neolithic and list the sites with skeleton remains.

What would be the best way to continue maintaining these archives in a flexible word editor (ex: Microsoft Word) while still being able to leverage AI/ML text extractive summarization for knowledge discovery?