word features for latent dirichlet allocation to extract the implicit aspects from the online reviews -- 2

I have implemented the code that required to extract the features "aspects" from the online reviews. the method that I have implemented is the Stick-breaking Process on the Latent Dirichlet allocation to extract the aspects"features" from the online reviews, But the problem is that while executing the algorithm, digital values are produced rather than "words" or features. in addition to that, I'm looking forward to integrating the "SenticNet 3" as an external knowledge to leverage the concepts to be used using the graph-based by adding a new layer to the term_topic_distribution as stated in this paper LINK of the paper "[login to view URL]". Where the SenticNet 3 should be used as an external knowledge to extract the aspects from the online reviews based on the above methodology. The attached files are the code that I have developed to extract the aspects without the implementation of external knowledge, I want the new code to be integrated into the existing code that I have attached. also, the attached MS word file is explaining the methodology of the wanted project. I want the code in python 3 and a clear explanation of the code that going to be included to my code that is attached.