Practical ways to reorganize decades of threads/content?

[haqzore]

Hi all. Would like thoughts / ideas on a change we're considering on a nearly 20 year old forum. I'll present a scenario that matches ours.

The forums subject matter is Italian Motorcycles.
For 20 years, we split a couple areas into subforums like this:

• Technical Info - Model Years 2000 & Prior
• Technical Info - Model Years 2001 & Later

We'd now like to split that first bucket, and and up with this:

• Technical Info - Model Years 1950 - 1975
• Technical Info - Model Years 1976 - 2000
• Technical Info - Model Years 2001 & Later

The challenge is sorting the existing 15,000 threads in the "2000 & Prior" forum. It would be supremely helpful to be able to identify the relevant 1950-1975 threads and move them into that new subforum.

One thought is to use AI to analyze, perhaps, the original post of all threads. I'd likely have to spend a good chunk of time drafting a prompt that gave accurate results. Similar to this idea - but that was a very if/then scenario vs a more analytical scenario like this one.


Does anyone have practical ideas on how to search, find & move threads that fit a certain criteria?
How / have you tackled a scenario like this in the past?
Appreciate any ideas.

 

[ehd]

If you let AI crawl your site for that, it will be quite difficult and AI messes up your urls. I did this with a similar task and all I got where threads with wrong IDs, which redirected the the thread with the ID.

One quick thing I tried today, was to upload all the sitemaps to the AI and let it work. Looks promising.

So I would just upload the zip with the sitemaps and add some of the directions you wrote in your post.

HTH

 

[haqzore]

If you let AI crawl your site for that, it will be quite difficult and AI messes up your urls. I did this with a similar task and all I got where threads with wrong IDs, which redirected the the thread with the ID.

I would hope to avoid just asking it to crawl the site - as I'd anticipate a lot of noise there.

I wonder if there's a way to export the original post + thread ID from a database? And that text could be scanned by AI?

 

[briansol]

AI is your best bet to get a first pass. it will still need moderation/manual efforts at some stage of the game.

searching for 1955 acme motorcycle won't give you the results you need. it will be missing years most of the time, '55 sometimes, using slang like fifty-deuce for 52 or whatever so things will get missed by any query.


generate a list of threads from that forumid so that ai only focuses on that (sitemap is too much).
have it crawl.
have it CLASSIFY the thread as pre 1950 or post 1950 subject matter.

models are really good at that.

take your pre-bucket it gives you and use that list to manually approve/move. leverage your moderators to help