Perhaps you might try LIWC (language inquiry and word count) machine learning algorithms.

Usually Random Forest does well on the whole, but it depends of the distribution of the response as well. 

Even I would suggest Random Forest algorithm.

I would agree, Random Forest would work well for your ensemble.

Thank you all.

However, if the dataset comes from a reservoir of web links, is it possible to apply the methods you suggest?

Turgay

What are the most accurate algorithms for ensemble learning with categorical dataset? Any algorithm to predict categorical items?

Best tool is Grenadian boosting machine(GBM) with cross valuation principle because it is not used the randomization concept such as Random forest,

Best Regards