20

Hello everyone,
I'm new here. [OK. Let's skip this]
I want to know where to begin on my journey on learning how to create a program that predicts what a user will say next by storing already said things and by making specific characteristics for the users.

I know that I will need to train it with some data first lol.
But how will it do the prediction. I just need this part of understanding.

I'm sorry for my bad English btw.

Comments
  • 3
    This might be interesting for you:
    https://en.m.wikipedia.org/wiki/...

    I used this algorithm once for a university project. Takes some time to understand it, but it could be helpful to you.
  • 2
    And welcome; have fun! :)
  • 3
    I think what you're looking for is a predictive model, now I don't know how experienced you are with the tags you've associated your rant with. But I'm going to assume you have at least a basic understanding of them all.

    Search up terms that I mention as whilst I can make a predictive model fairly quickly, I'm not all that good at explaining things.

    You should do some form of analysis by building models based in logistic regression and decision trees. You can analyse a lot of data quickly with some modern machine learning tools. That will form the basis for your predictive model.

    Your data will probably be very messy after you've done your analysis, try to remove outliers and fill in missing values with averages around it if you want to cheat and treat your data to manipulate it into giving you some quicker results.

    There are some techniques you can look up to make your model more accurate, you could use gradient boosting algorithms, honestly when I did messing around in R that....
  • 3
    ... is all I used when doing predictive models. After that you can whip out your training data and whatever test data you want to try on your new model. Lots of Googling to be had but if you have any questions just comment with an @SoulSkrix tag and I'll reply when I get a chance :)
  • 1
    Recurrent Neural Networks for you!
  • 1
    You can actually get really far really easy if you use Hidden Markov Models with bigrams and trigrams. That should be your starting point for text predictions. When you get the hang of that you can mix in neural networks and other advanced stuff.
    Just write a simple HMM program and let it 'read' some books using trigrams. After that you have a simple autocorrect/autocomplete keyboard.
    If you want to predict on higher levels than just words, let's say on sentences, you can actually use these same trigram and bigram models mixed with some cleverness of your own and a more advanced learning algorithm such as neural network stuff.
  • 2
    No clue but welcome!
  • 3
    Thanks for all the greetings :)
    And for the helpful answers of course. I will see what I can figure out.

    This place is so awesome. Everyone is so nice to everyone :3
  • 2
    @teganburns arigatou gozaimasu.
    I like her lips
    Js lmao
Your Job Suck?
Get a Better Job
Add Comment