Narrative Focus Indices

[mas-4]

I think we should create a Story Table that monitors ongoing political narratives. Basically a headline gets added to a story if it reaches some threshold of words that match the story. For instance, consider the following:

1. Biden Age: [biden, age, old, memory, senile]
2. Trump Unfit: [trump, fitness, office, dangerous, allies]
3. Guns: [shooting, mass, weapons, ar-15, ar15, guns, dead, killed, injured, shooter, control, 2nd amendment, second amendment]
4. border: [illegal immigrant, migrant, crisis, border]

This is wrong and I just talked to chatgpt about it. It seems topic modelling is a well defined concept in ml and I need to investigate it. My only real worry is how much processing power this is going to take. I wonder if I'm going to need to enlist my 3080 which is not something I like.

I could eliminate the articles all together and speed up the process of everything. Articles are really never read and barely influence public perception. Its worth noting this whole exercise is meant to be a realization of Stancilism, Will Stancil being the plucky tweeter that argues media narratives control public perceptions more than people think. So conceivably we should abandon articles and try to do a kind of moving windowed topic modeling. There come to be a couple of parameters we need to pay attention to:

The current topic window is most important. Say a month long window.

But there are continuing topics (Biden Age, Guns, Border Crisis) that have lasted years now. GWOT is no longer a real topic though, China Threat is!

So how we balance all this I'll need to explore by just doing some basic topic modelling to see how the process works. But this is largely key to my interests. Paying attention to what stories get picked up and how much coverage is devoted to them will largely determine the outcome of the election, sadly. So knowing which narratives are high on the list will be essential to my process.

I might also want to create a blog on this site.

[mas-4]

I honestly think we just need to use keywords here but i'm increasingly wanting also to group headlines for a given story.

I think I should check out k means clustering but i also think that I can do the following two things:

1. Identifying a story (over a narrative) can be done by creating a flask admin panel that allows me to pin a story as exemplar and then use cosine similarity or something similar to identify matching stories
2. Election Narratives can be tracked just by keywords