RE: [Steem Rep] Update - September 2024 | AI-Comments | Tags | Trendings Scores
Nice to hear your thoughts - I looked yesterday to see if you'd shared anything and today, you've fulfilled my wish (although it was probably yesterday after I'd looked for you).
AI Generated Content
I've struggled to know what to write about this. I also wanted to share these thoughts privately before commenting publicly.
Obviously, I don't share the opinion of the user who was trying to incite a new wave of Witness Wars. In my opinion, AI generated content which hasn't explicitly stated that it's AI generated deserves the downvotes it receives - irrespective of whether it's a new user (in which case, more leniency would be preferred), or an admin/moderator or even more so a Witness who have the responsibility of taking this platform in the direction that we want it to go in. We must remember that this is why we vote for witnesses - to support the users that will take the platform in the direction that we want it to go in.
In my opinion, xpilar failed in this respect and thanks to @michelangelo3 who highlighted this comment that has done a good job of combining many of my feelings in a relatively short comment (surprise and disappointment being front and fore).
Before this user reappears with another wave of blind loyalty, I have to reiterate my admiration for xpilar and everything he's done for Steemit, its users and for me, personally (which is why he still gets my Witness vote). This drama was in my view, poorly judged and I suspect, poor advice from elsehwere.
Tags
Yep - URLs matter, quite a lot. And the earlier in the URL that a key term appears, the more importance a Search Engine has historically given it.
So having hive-xxxx - a meaningless search term appearing in the URL before the h1 is essentially saying that hive-xxxx is a better descriptor of the content.
When I finally have time to write some new content, my intention is to use SEO optimised accounts (with a view to promotion of the platform to new users and how to join, etc.) and then using my main account (this one) to say "I wrote this - if you like my content then please take a look". This will also have a doubly-positive impact that if my main account's post exceeds $20, there will be an additional internal link to my article which is also SEO optimised. Win-win. At very least, this is the theory which I can test and potentially fail.
The SEO also links to your next point about the Trending page - This page is the one that Google ranks highest on steemit.com - https://www.google.co.uk/search?q=site%3Asteemit.com and is therefore the one that it will crawl the most regularly. If the content which appears regularly at the top of this is the content that it's following and indexing, this doesn't bode well for the quality of content that Google's indexing. So, the algorithm is a crucial development in my opinion.
Trending Algorithm
Both of your suggestions are good in my opinion. I think the Trending page would benefit if voting bot votes are removed (without thinking more complex - this change would see a huge improvement) and potentially all auto-votes (although these are sometimes a good indicator of the better authors - although maybe not. Some authors who receive over 1,000 votes per post are crap).
This does then mean that votes from sc01 become all-powerful in topping Trending which (in my opinion) becomes a problem when "Steem POD" posts get the highest upvotes. Is that the type of content that will appeal to new users or should consideration be given to how sc01 and sc02 votes are also handled within the algorithm?
You are of course particularly affected at WOX, as it affects the founder (and/or his team)... and perhaps this indirectly impacts the community mods as well.
I am very amazed that the comment highlighted by michelangelo3 was written. And furthermore, I'm surprised that no one has responded to it yet. I agree with you that exactly the right words were used.
I don't know what you mean by that. There is no
h1tag in the URL.Yes, but doesn't the lower SEO of the main account then influence the SEO optimised account? Or is the link alone a criterion for a ranking increase?
Unfortunately, it is not possible to determine whether the vote has been assigned manually or automatically. Unless we define very early votes as autovotes. As I wrote, I am not (or no longer) so convinced that bot votes alone will help us, as many people follow these bots. But until I've done the maths, we're only talking about it in theory. I'd like to know the actual impact so I can weigh up the alternatives. Which brings us to the most difficult point: How do I make comparative calculations without having to change the whole Hivemind code... I'm still thinking about it :-)
In any case. These votes should not be the decisive factor, but the broad masses. In my view, remlaps' suggestions are (theoretically) well suited to achieving this.
Sorry, I'm partially referencing michelangelo3's comment. The "Title" that we use for our post is also used as the
h1tag, as well as to construct the page URL (unless the title is edited,h1= page URL). So when the page URL is constructed, the community (or first) tag precedes the title which would be given slightly higher importance.I might find this difficult to articulate but I'll do my best 🙂 If done well, the 2 should compliment each other. If the main account says something along the lines of "I have written this article about Lego Set 14433 which details my experiences building the set, and how much I enjoyed playing with it, blah blah" and then links to
Lego Set 14433 - Building and Play Experiencethen when Google looks at the refferer content and deeper content, it'll say "Yes, these are relevant to each other and I like this". Whereas if I dumpLego Set 14433 - Building and Play Experienceat the end of a "Gareth Southgate depresses me" article, it will say "Huh? What's lego got to do with Gareth Southgate - this fool's link spamming, I don't like this".In the first (honest) example, Google will like both accounts because they're acting honestly (of course, this is all algorithmic) whereas in the second (dishonest) example, at minimum, the linking account is penalised and depending upon domain usage, both accounts might be. Internal links would be weighted differently to external links - another example is if at the end of my "Lego building and play experience" article, I link to other relevant information (perhaps the set on Lego's website), Google would also like this (assuming that the post reaches $20 and the
nofollowlink is removed.Overall, the consistency will also appease Google. The account "Lego-Builder" who always talks about Lego will always rank higher than a generic account like mine which talks about lots of things. Google will have more confidence that the results its displaying will be relevant to the search term.
So lots of factors required to make the idea work and ensure that Google doesn't see it as "This fella's trying to manipulate me" which Google's been working very hard to prevent and dislikes. Which is why their algorithm's always been so secretive - if people know their rules, they'll play the game.
This is what I was thinking. At the moment, we know that autovotes tend to happen after about 5 minutes and sooner for the biggest voting bot delegations. Of course, if the rules change, voting habits probably will too.
In the main, I'd say these followers have relatively low voting power with less influence.
I think that we can say with a high level of confidence that if voting bot votes were removed, it would instantly signify a massive improvement in the Trending page. It might also highlight the next challenge (whether that's the voting bot followers or something else).
I'd be tempted to approach this as incremental changes rather than trying to find the silver bullet straight out of the gates. Of course, using suggestions by remlaps - I know that he and others have thought about this a lot over the years.
Thanks for the detailed explanation. :-)
Now I can understand it better.
I don't know if step-by-step is appropriate here. I don't want to find the ideal solution straight away, but I would like to compare how the different approaches rank.
The trending algorithm is calculated continuously (with every vote) and not just when the trending posts are requested. In this respect, changes to the trending page do not occur immediately when the code is changed. So you have to wait a few days until you see other posts on the trending page.
I had therefore considered creating a separate table in which the trending scores of the various calculations are saved so that a direct comparison could be made. For the display, however, we would then have the problem that there would also need to be a selection option for the respective trending algorithm in the Condenser. Hm, that would be more work than necessary.
When I recapitulate these thoughts now, I am inclined to simply try it with an algorithm and then see what changes in the trendings. We can then at least make a live comparison with the Steemit variant.
Could you please send me your bot list via Discord so that I can use it?
!DUBby 18%
Done and done. I do like the idea of running lots of algorithms and storing additional columns of data (which would allow a relatively quick switch (in the code) to see different attempts). Overall, this might be quicker than waiting a week at a time to implement different possible approaches.
I don't know how easy it is to add an additional database column with associated ranking algorithm though. It's one thing to think that it sounds straightforward enough but very different to implement it on Steemit's codebase 😆
Before I start anything else, I wanted to finish the new version of DUBby first. That's why I've only really had time for these things for a few days... Although I was already thinking about it :-)
Today I had to struggle with my inkjet printer for a few hours to get it to spit out one of my daughter's school homework. Of course, by the time you need it, something has dried up...
So I'm sorry I'll have to postpone replying to your comment and the others...