I have a sort-of similar goal with my research data bases in Tinderbox. I accumulate notes over time on subjects I'm looking into. Then, when it comes time to put together an article or book or speech, I want to be able to dig back into this stew of notes and find connections--both connections I saw and intentionally built in as the project was underway, and ones that emerged.
I have three main ways of doing this, all relatively simple in tech/coding terms:
Tags and themes. My main research data-bases have a very simple structure. After every interview or research trip or whatever, I put the substance of the data in the text of the note -- and then fill in a few attributes. $Source (who I am interviewing, or what I am reading); $Date of course; $Location (Washington, Shanghai, phone interview); and $Tags. These tags range from general -- China, military, politics -- to specific (carbon capture, F-35). Then when I'm beginning to write or plan an article, I almost always use the attribute browser to see all the research I've collected on carbon capture, or all the interviews I did in Sioux Falls, South Dakota. This thread has lots of illustrations of using the attribute browser -- for my money, the most powerful single tool in TB.
Maps and links. Some people use these very extensively, as a way of making, revising, and envisioning connections among items. For my style of thinking, I use them somewhat less -- but still find them valuable when doing large-scale planning for a project. Eg my wife and I have been finishing a book recently, and we used several big-scale TB maps as a way of figuring out (and revising) the sequence of our respective, alternating chapters.
Detailed article-planning tools. Usually when getting ready for actual writing of a chapter or article, I'll go back through my main research notes to see, in a much more precise way, the quotes and points and citations I want to use. Then I have a working-draft data base where the main unit of info is the specific point, quote, or reference I want to use -- these range in size from just one fact, phrase, or sentence to a paragraph or two.
Then I tag these in a more complicated way: which chapter or section of the article I want to use them in (so I can see, in the attribute browser, the relevant info as I'm doing the writing); a checkbox for when I actually use them, so i don't use them twice; whether they're "Hot," so I have to be sure to include them; what the source was, so I can see all quotes from Mr. X when I call him for fact checking; and what theme the item supports. On this last point I have an attribute called $Themes, set-style, where I can give each item one, two, or very many descriptions for the points it might make. "New role of downtowns"; "paralysis of national politics"; "evolution of career-tech education"; and on and on. My conception of what will be the appropriate thematic tags emerges as I'm going through the material, and as I tag a bunch of items I can see in an attribute browser view what themes and connections are developing that I hadn't fully anticipated, how to connect them, etc.
In practical terms, when using these detailed notes for writing, I usually have one main TB window (in previous versions of TB had several windows), with a large number of standing Tabs, each of which has a relevant view of the data. One will have an attrib-browser view of the notes arranged by chapter or section of the article. (It will include columns, so I can see the source and theme -- and checkboxes for when I've used the material.) Another will be attrib-browser view by theme; another by source; etc.
Typing this in haste, so forgive typos -- but it's an illustration of the way the tool can evolve to reflect your interests and needs.