This is a sample project developed in python and heroku. It analyzes product reviews at Amazon to extract most frequent words in various product categories to determine important qualities that users care about most. Such data will be valuable for business owners and for people looking to open new businesses in that area.
It uses Flask for html rendering and bokeh for interactive plots. The word cloud is generated by the javascript library JQCloud and JQuerry. The sample data is amazon product reviews for 20 tablets downloaded from data. hosted at http://fast-shelf-85324.herokuapp.com/
Generates a word cloud from the amazon reviews for the tablets product category. Upon clicking Generate WC, the app reads a json file containing the data, cleans up the reviews, removes common words and calculates ranks based on frequency. The results for top 30 words are displayed using JQCloud. Note that it may take a while. However, a pre-calculated image can be displayed by clicking Show WC.
Not surpricingly, the most important quality for tablets is "app".
For specific product (in this case Apple iPad), the frequency of few of the significant words can be plotted as a function of the time, to see if their important changes over time. In this case, the code calculates word frequencies and corresponding times for every review, occurance of words at various times and sorts the results. It is then plotted using bokeh library. Plots for "app", "price" and "screen" can be selected below.
One can see that "app" was an important property when iPad was launched, but has declined over time. However, "price" and "screen" remain more or less significant.