Raw code written and executed at Google Colab :
FOR END-EVALUATION : https://colab.research.google.com/drive/1MVInaCIV1e6-wl8eQYNFZxDdpwdni7Tz#scrollTo=4I1UOMg9Lpo7
SECONDARY GOALS : https://colab.research.google.com/drive/1rZsGGzT0ODys4J72z4R8XwwISMivBYq_#scrollTo=8Uwn4Lte_eEM
I have finished the code for LDA model, trained the model, tuned the model, calculated the Coherence Score ( it came out to be 0.39888 compared to the optimal value of 0.454) and also I finished the part of secondary goal which was the sentiment analysis of the reviews. I have created Flask Interface for Secondary goals part but it might not display any rating beacuse of some bug & I could not correct it as of the time of submission but I will surely fix the bug asap & will merge the two files to create a final model which can show both the proportion of words in the review as well as the nature of the review( +ve/-ve ). My Flask API Link : http://127.0.0.1:5000 I have attached all the files I created & used.
How to interpret the LDA model ?
The given image shows the output of an LDA model which shows the AVERAGE TOPIC COHERENCE along with the tokens(words) with their proportions in the document we created in LDA.
The image below shows the visualisation of the topics created using the python pyLDAvis library . Just move the cursor on the topics shown in circle to see the frequency of the different terms.
Happy Topic Modelling :). I hope now analysing the reviews become much easier for you compared to what it was before.