I would like to create an NPC engine for games that are in a fantasy context like Lord of the Rings, Warcraft, Skyrim, Dragon Age, etc.

This engine should be able to determine the polarity of the given sentence (positive/negative/neutral) and should also be able to tag the words (part-of-speech tagging, but that's not relevant now). The problem is that I cannot find a dataset which I can teach my AI with, so that fits into the fantasy context. As an example for Lord of the Rings, there are characters like Sauron which are negative, but there's a ton of expressions too in a fantasy world which cannot be found in a general dataset.

Therefore, I would like to ask if you could suggest taking dataset(s) which contain expressions, even better names in a fantasy context. It doesn't matter which fantasy universe does it take.

Thank you!

  • 1
    $\begingroup$ Sounds like an interesting project. Sadly, the more interesting and original your project idea, the more likely you are going to need to source your own data the hard way. You don't make it clear, but I assume you want sentences tagged with sentiment labels from those fantasy settings, not just the text from some novels (which would be a relatively easy request, provided you avoid copyright issues)? $\endgroup$ – Neil Slater Oct 30 '18 at 11:20
  • 1
    $\begingroup$ Yes, it's something similar. First, I need sentences and expressions with their polarity. I can implement tagging by using the language rules. One of my friends suggested movie subtitles, but the biggest problem with them is that they are relatively short for a dataset (even if I put together more). I also need to specify the polarity of each sentence and expression manually. $\endgroup$ – NagyDani Oct 30 '18 at 12:11
  • $\begingroup$ The data would not be available so easily. But you can use Google's Dataset Search tool. It is very useful and has many datasets indexed. You can know more about it from this official blog. $\endgroup$ – Shubham Panchal Nov 3 '18 at 2:20
  • $\begingroup$ Have you evaluated existing data sets on this? Given that there accuracy is often not all that great anyway, it might not make too much of a difference in the fantasy domain. Also, often you have word lists annotated with polarity, which you might be able to adapt. $\endgroup$ – Oliver Mason Nov 16 '18 at 16:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.