12 {{ upvoteCount | shortNum }}
12 {{ upvoteCount | shortNum }}

Markov chain blogging?

by CDarklock
10 replies
Rather a technical question, I know, but is anyone here using Markov chains for autoblogging?
#blogging #chain #markov
  • Profile picture of the author Matt Bard
    I don't understand why Markov over just good old fashioned "random".

    What would the benefit be?

    Matt
    {{ DiscussionBoard.errors[1551703].message }}
    • Profile picture of the author xiaophil
      Originally Posted by Matt Maiden View Post

      I don't understand why Markov over just good old fashioned "random".

      What would the benefit be?

      Matt
      Texts generated using Markov chains are not purely random, and tend to stick to a topic - sometimes almost passing for a real person.

      I suspect some WF members might be Markov text generators.

      What's the point though? While they can certainly generate unique content, it's not usually suitable for human consumption.

      Phil
      {{ DiscussionBoard.errors[1551796].message }}
      • Profile picture of the author CDarklock
        Originally Posted by xiaophil View Post

        Texts generated using Markov chains are not purely random, and tend to stick to a topic - sometimes almost passing for a real person.
        One of my Markov generators has been published in three anthologies of poetry. Of course, this rather begs the question of whether poetry editors are in fact human, but all the same... it's something I've toyed with off and on for several years, and with an appropriately focused corpus, I think the results might be better than you expect.

        Because autoblogging is kind of like throwing a spam filter into "reverse"...
        Signature
        "The Golden Town is the Golden Town no longer. They have sold their pillars for brass and their temples for money, they have made coins out of their golden doors. It is become a dark town full of trouble, there is no ease in its streets, beauty has left it and the old songs are gone." - Lord Dunsany, The Messengers
        {{ DiscussionBoard.errors[1551910].message }}
        • Profile picture of the author xiaophil
          Originally Posted by CDarklock View Post

          One of my Markov generators has been published in three anthologies of poetry. Of course, this rather begs the question of whether poetry editors are in fact human, but all the same... it's something I've toyed with off and on for several years, and with an appropriately focused corpus, I think the results might be better than you expect.
          I'm aware of some famous poetic examples, and a couple of famous chat-bots, but kind of assumed with blogging you were leaning more towards articles. While it would be great for a poetry or musical lyrics blog, I've never seen anything resembling a coherent article generated purely from Markov chains, even with a large corpus.

          One of the major issues I think is that regardless of the n-gram length, the most likely choice of word is still often wrong, and errors are cumulative for at least the sentence. I would however be happy (and quite amused) to see one perform otherwise. Perhaps a little heuristic guidance or something?

          I think there are plenty of opportunities for auto-blogging using statistical methods and also computational linguistics, that would not only be light years ahead of anything that's currently publicized, but also generate genuinely useful content.

          For one thing, very little is being done in auto-blogging ( that I can see anyway) with relatively straightforward techniques such as naive Bayes classification, which could raise the quality of a site immensely.

          I've been toying around with things like paraphrasing texts using bidirectional translation against the Europarl (European Parliament) parallel corpus. Nothing new as far as techniques goes but a very interesting baseline. Pretty resource hungry though, in more ways than one, I have to be careful not to get sidetracked. Fascinating work but it doesn't (yet) pay the bills.

          Another great way I think would be to generate content directly, nicely formatted, from otherwise invisible or hard to access databases. There's a few I notice around doing this, and they seem to be doing well.

          Because autoblogging is kind of like throwing a spam filter into "reverse"...
          I'm not sure, wouldn't that mean you put a little bit of good stuff in, and then get a ton of crap out the end? Hmm.

          Phil
          {{ DiscussionBoard.errors[1552319].message }}
          • Profile picture of the author CDarklock
            Originally Posted by xiaophil View Post

            While it would be great for a poetry or musical lyrics blog, I've never seen anything resembling a coherent article generated purely from Markov chains, even with a large corpus.
            Well, on my side of things, I've seen an awful lot of successful blogs out there which don't have anything resembling a coherent article...

            I'm not sure, wouldn't that mean you put a little bit of good stuff in, and then get a ton of crap out the end? Hmm.
            No, that's called an article spinner.
            Signature
            "The Golden Town is the Golden Town no longer. They have sold their pillars for brass and their temples for money, they have made coins out of their golden doors. It is become a dark town full of trouble, there is no ease in its streets, beauty has left it and the old songs are gone." - Lord Dunsany, The Messengers
            {{ DiscussionBoard.errors[1552484].message }}
          • Profile picture of the author bgmacaw
            Originally Posted by xiaophil View Post

            I've never seen anything resembling a coherent article generated purely from Markov chains, even with a large corpus.
            My Blog Content Wizard program occasionally amazes me with some rather lucid passages, for example, this passage constructed from blog content keywords...

            Blog content may entertain us or make us think twice but you might think that I'm talking out of my rear end. These are affluent times for autoblogging owners. It is well that it should be so. You could make this a livelihood. This should fan the flames. I probably should be more diligent with using that although a lot of specialists do know how to make it big with blog content. I hadn't believed that I should provide more details. Well, feast your eyes on this. You don't have to be brainy however I'm sitting on the fence. It isn't a time to wing it. I'm not going to have pals working against me on this. I've seen quite a few blog content analysis and none are close to this. You'll want to have them in your pocket. It's as slow as molasses. I've assembled a team of experts on this. Few cooperatives realize just how powerful doing this is. It doesn't require any technical ability so that now we're on easy street. Blog content is easily overlooked. Who would a thunk it, huh? You are going to need to make sure that you know what sort of blog content idea you want to get.
            You are correct that the underlying dictionary and seeding has to be very big and the various elements used have to mesh well together. Otherwise it isn't too coherent, such as this more typical passage on digital cameras...

            Women like to read this touching on digital camera. It is my professional data provided that it is really duplication proof. It is advised by consumer report digital camera experts. To what degree do top dogs come upon luxury pocket digital cameras steps? Probably not, unless you discover that it works for you yet let's say it's about doing that. I don't have to be deflective. It is one of the toughest things I have found. Some of the variables which make shopping for this contraption so difficult include the following things. "No problems"! Who are they kidding? A mind is a terrible matter to waste and this, we are told, was somehow a bad thing. I have been trying to find a good digital camera ratings company for a while now. If you thought that was difficult, try this. Here's how to stop worrying about what other associations think of you. Let's take on cheap digital camera first so that there's no one here but us chickens. You need to take small steps at a time.
            Signature
            Product Reviews | Earn Online Cash | Free HTML Templates
            Free WordPress Themes: Boring Memo | Dateless Mini-Site | Info Magazine | 100 Twenty-Ten Niche Headers
            Discount Templates, Graphics and Scripts: Templates for Website
            {{ DiscussionBoard.errors[1553816].message }}
  • Profile picture of the author ildarius
    how about Ovechkin chain blogging, or Kovalev for that matter
    {{ DiscussionBoard.errors[1552824].message }}
    • Profile picture of the author CDarklock
      Originally Posted by ildarius View Post

      how about Ovechkin chain blogging, or Kovalev for that matter
      A stochastic process has the Markov property if the conditional probability distribution of future states of the process depend only upon the present state and a fixed number of past states; that is, future states are conditionally independent of past states older than a fixed number of past states.

      What are the Ovechkin and Kovalev properties?

      You know, other than hockey uniforms.
      Signature
      "The Golden Town is the Golden Town no longer. They have sold their pillars for brass and their temples for money, they have made coins out of their golden doors. It is become a dark town full of trouble, there is no ease in its streets, beauty has left it and the old songs are gone." - Lord Dunsany, The Messengers
      {{ DiscussionBoard.errors[1553354].message }}
      • Profile picture of the author ildarius
        Originally Posted by CDarklock View Post

        A stochastic process has the Markov property if the conditional probability distribution of future states of the process depend only upon the present state and a fixed number of past states; that is, future states are conditionally independent of past states older than a fixed number of past states.

        What are the Ovechkin and Kovalev properties?

        You know, other than hockey uniforms.
        None that come to mind apart from high salaries, fame, loads of hot female fans and absolutely no need for auto-blogging

        Personally had a network of 500 blogs with "re-translated" content, on a free blogging platform. After making around 30$ per day in adsense some blogs started getting 0 traffic while the rest got banned. Surprised the adsense account wasn't cancelled altogether.

        In terms of link juice, didn't see any significant difference in rankings when linking out of those blogs.

        The content on those blogs read like a 1$ 3rd world country article.
        {{ DiscussionBoard.errors[1560231].message }}
  • Profile picture of the author Karomesis
    Have you researched MOEA? (multi objective evolutionary algorithms)

    I've found them to be very useful across a wide range of industries and markets.

    There are a few white papers on using them for IM purposes, I think i saw them in a marketing journal somewhere, I'll see if I can find them and post in this thread.
    Signature

    Coming soon....FULL SCALE AUTOMATION.
    "Set it...Forget it" site building and SEO software.

    any ? please hit me up anytime karomesis12@gmail.com

    {{ DiscussionBoard.errors[1560888].message }}

Trending Topics