The Altair Community is migrating to a new platform to provide a better experience for you. In preparation for the migration, the Altair Community is on read-only mode from October 28 - November 6, 2024. Technical support via cases will continue to work as is. For any urgent requests from Students/Faculty members, please submit the form linked here

"How does Sentiment Analysis accelerator creates continuous variables?"

NickMNickM Member Posts: 2 Contributor I
edited June 2019 in Help
Hi all,

For my MSc Thesis I am delving deeper in predition analysis. I know that Rapidminer has an accelerator for Sentiment Analysis, but I am wondering how they create the scores. If I understand well, a SVM is used on the dataset that you enter. However, the result of a SVM was to create a hyperplane that seperates points +1 from -1 right? So how are the continuous variables in the accelerator created then, and are they actually valid?
Tagged:

Answers

  • David_ADavid_A Administrator, Moderator, Employee-RapidMiner, RMResearcher, Member Posts: 297 RM Research
    You can check out the underlying process of the accelerator.
    In the results view there is on the lower right side a button "Show the process".

    Regards,
    David
  • NickMNickM Member Posts: 2 Contributor I
    Thanks David, but it's exactly the process that I'm not relying, so I was hoping that someone could explain the process in normal terms. :-) When you look at academic research regarding SVM, I've not yet found any paper that converts the binary output to a continuous one. That's why I'm not really trusting the process of this accelerator.

  • David_ADavid_A Administrator, Moderator, Employee-RapidMiner, RMResearcher, Member Posts: 297 RM Research
    You'r right, the classic SVM is only capable of seperating between two types of values (like +/-1).
    What is done in the accelerator is, that the confidence values for a positive rating are transformed into an integer value. These values originate from how far a value is away from the separating hyperplane, the greater the distance to the hyperplane, the lesser is the confidince that this value is classified correctly.
    The transformations are done in the "Best and Worst" subprocess.

    So there is no real magic behind the continuous output of the accelerator and no reason to distrust the results  ;)
Sign In or Register to comment.