Comment generation thesis

Discussion in 'Education' started by nickdesmedt, Dec 25, 2017.

  1. Hello, my name is Nick De Smedt. I'm a student at the University of Ghent, Belgium. For my master thesis, I've created a picture comment generator. To test this, I've set up a platform that contains 50 picture-comment combinations. Could you please take a couple of minutes of your time and rate these comments?

    This is the link: http://193.190.127.170:4242/

    Please note that the comments are heavily preprocessed (capitals and punctuation removed) and might contain abbreviations/technical terms. If you have any questions, feel free to contact me here or at npdsmedt.desmedt@ugent.be

    Kind regards,

    Nick De Smedt
     
  2. While I sometimes think that some comments here are automated, I'm not exactly clear WHY you are doing this.

    And I am NOT going to click on your site.

    Welcome to Photo.net, I guess.
     
  3. Frankly, I doubt you will get many responses without a detailed explanation, and even then, most prudent individuals avoid strange links. Possibly some other venue or approach may prove more effective. Good Luck with your project.
     
  4. I'm doing this for scientific research purposes. I need people to rate these comments to see if the computer generated ones are clearly distuingishable from the human ones.
     
  5. Science tells me and others not to click on non-secure and non-traditional IP addresses from strangers on the internet.
     
    scottroberts likes this.
  6. Yeah, well....
     
  7. I think I'll pass...
     
  8. For the nay-sayers: the email address posted is consistent with the university mentioned, and the user's name is typical for the region too. So the sceptism, while logical (see below), may for once be misplaced....

    YET....
    Nick, there have been requests like this multiple times on this site, and quite often either click-bait or outright phishing attempts. So it's very logical people are suspicious, and a link to an IP address with no means of validating its validity is a great way to harvest some malware. Likewise, if your webserver is a legit thing, posting its address on a website is a great way to attract unwanted traffic, so I hope you have considered your security.

    Furthermore, it would already help if you'd be more clear about the purpose of your research: instead of simplifying it to "distinguish between computer generated comments and human comments", how about a bit of an idea what that yields scientifically? Are the computer generated comments based on image analysis using large datasets to use determine which comment to use? What are you actually trying to prove? It could help people persuade participating. Just saying it's scientific and asking people to risk their computer safety is a bit a stretch.

    So, yes, count me in with those that do not click the link. Hopefully the above comment however could give a bit an idea of how you could easier persuade people to do participate instead.
     
  9. Shouldn’t PN administration play a role here? Shouldn’t there be some way of monitoring such posts and either verifying them or deleting them in order to ensure the security of membership? Maybe that’s asking too much. I really don’t know. But I, too, wouldn’t click until I knew more.
     
    scottroberts likes this.
  10. Wouter has it absolutely right. Everyone these days has been bit by supposed "research" that turns out badly.
     
  11. I'm not interested in anyone who wants to see if a computer response is as good as a humans'. I hate the damn things, and think such research is a bad idea...
     
    Jochen likes this.
  12. Science isn't a bunch of answers to a bunch of random hypotheses: "See if this happens, see if that happens, etc." For a question or hypothesis to be meaningful, it has to be part of some larger structure or theory. I don't know if you'll have any success in getting photo.net member to help you, but you might try by explaining what you are doing more fully or providing references to published materials.
     
  13. As someone who has an advanced degree in a "hard" science, it often astounds me how many folks I work with who just completely forget the scientific method in their research.

    Just as a refresher-

    1. Make an observation

    2. Formulate questions based on observation

    3. Generate a hypothesis that might explain those questions

    4. Design an experiment to test the hypothesis

    5. Evaluate the data from the experiment to see if it is consistent with the hypothesis

    So, I'll go back to the fundamentals and ask what observation led to the eventual design of this experiment.
     
  14. To me that's easy. Picture comments are first and foremost should be only positive, nothing negative. Second the comment should have no mention of anything in the contents of the pictures. People would like that and of course they are all useless.
     
  15. Sorry, the survey exceeded my attention span and I didn't finish. - If I post pictures online, I'll get stupid comments (at least outside this nice community). Do I need artificial intelligence to generate those? - IDK. - I am sure real AI should be able to generate something helpful by now. - But why should it try to impose a human?
     
  16. I'm among those who chose not to click on the link. From a theoretical perspective, I have very serious doubts as to both the desirability and plausibility of meaningful critiques generated by AI. When I choose to critique an image I employ all of the very highest intellectual operations, on both right and left sides of my brain. Then, I very carefully select the language, both words and tone, I will use to communicate my thoughts. Until we have an "emotion chip" that works at least as well as Data's (Star Trek reference, for all you non-nerds), I don't see how a computer generated critique will be nearly as meaningful as those offered by thoughtful, knowledgeable, and considerate people.

    I'm reminded (again) of my wife's explanation of art to our daughters: When asked what something "means", she explained that it is not so much about meaning as it is about how the art makes one feel. Until a machine has feelings, this all appears a moot point.
     
  17. I think an alternative can be offered so that the importance of feelings in art is communicated but not at the expense of meaning.

    Art can be and often is very much about and dependent on meaning and meaning doesn't, IMO, necessarily play second fiddle to feelings. What would a Shakespeare play be to me without the import of meaning? I can't imagine. Poetry is sound, music, and meaning, isn't it? How could symbolism, so vital to so much art, operate without meaning? How do I feel anything for the goings on in a film like Schindler's List if I don't get the meaning behind it? Much art is as important for the meanings it conveys as for the feelings it generates. Sometimes the meaning is actually quite a bit more important than the feelings involved. "What does this mean?" seems to me a very important question to ask of any work of art.
     
  18. In reference to this topic that would have to be determined by the AI within the computer algorithm. I don't see what alternative POV's about feelings and meanings purported within the creative process has to do with artificial intelligence unless just for the sake of arguing.

    I think all of us here can figure out artificial intelligence has a long way to go before it can come across as having real and genuine empathy, feelings and meaningfulness.
     

Share This Page