To classify text I would prefer fast statistical classifiers (Bayes, Winnow etc.) that are used in spam detection. They take words (or combinations of words) as a token and assign probabilities to them. The probabilities of all tokens of a text determine the class. Training of statistical classifiers is much faster than neural networks.

Could you perhaps clarify what type of labels you are going for?

Some time ago, I worked on a project for sentiment analysis: determine whether a text had a positive or negative connotation. For that project, one of our main cues was the occurrence of certain keywords; in theory, you could apply the same technique to a neural network classifier as well. However, other classifiers would seem more appropriate to me. Unless you can find a small number of high-entropy features, your classifier will suffer from the curse of dimensionality. 

I suggest you to look into spam classification. That will get you a few ideas how to get started.

Hi,

You must properly label the text or its sentences based on the importance and train the neural network. For identifying the important features in the text, I advise you to skim through the journal below

https://digital.library.txstate.edu/bitstream/handle/10877/3819/fulltext.pdf

you may explore models like SOM and ART for solving your problem.

To classify text I would prefer fast statistical classifiers (Bayes, Winnow etc.) that are used in spam detection. They take words (or combinations of words) as a token and assign probabilities to them. The probabilities of all tokens of a text determine the class. Training of statistical classifiers is much faster than neural networks.

There are one additional approach to quantify text: usage of Osgoods Semantic differential. This approach allow to avoid the dimensionality corse.

Hi Anoop,

Perhaps you could define what are the labels that you're looking for in the text. Is it their frequency that you are trying to pin down. Or other sort of classification? Have a look at this:

http://textblob.readthedocs.org/en/latest/classifiers.html

Good luck! Luiza

How to assign weight or decide on hidden layer in NN?

This might help you: https://radimrehurek.com/gensim/models/doc2vec.html

I recommend you try Rapidminer;