Sentiment Analysis for local languages (Nepali)

I would like to do sentiment analysis on document level. But I am try to do sentiment analysis Nepali. So, I dont have any resources. I can't do Naive Bayes Classifier as I don't have any labelled data and I can't do vai wordnet as no nepali wordnet exist. Papers I read generally had labelled data or senti-wordnet for other languages.

I would like know these things:

• Which approach should I use in above case for sentiment analysis?
• Is there any method for me to dynamically generate labels for data?

Since you don't have any labelled data, Have a look at this GitHub Repo, feel free to fork.

It has the code for neural network for Handwriting recognition in Java. Jeff Heaton has done it easy for us, with a nice UI, you can train this model to recognize Nepali.

And for sentiment Analysis, you can try using Opennlp which has some good support, this blog for Beginner's.

Also DL4J is a good library for deep learing for Java which can be used for Sentiment Analysis. It has a good Word2Vector Implementation and has a lot of support.

These resources will help you, any futher doubts-feel free to comment.