LDA Topics
Below is an interactive model to explore the topics I discovered using gensim's LDA topic analysis. I have separated user reviews into 8 topics - you can explore these by either clicking on the bubbles, or entering a number into the "Selected Topic" filter on the top right. The size of each bubble represents the number of unique words attributed to each topic, and you can consider that bubbles closer to each other share more in common than those further apart.
On the right, you will see a list of words that best define each topic (if you're a statistics person, see the equations at the bottom of the visual). The full bar represents the frequency of words throughout the entire dictionary, while the red highlights represent the frequency within the topic. In cases where the red portion fills up nearly the entire bar, this word lives almost exclusively within the chosen topic.
The visual defaults to a relevancy of 1.0, which means that it looks at the most frequent words in each topic, regardless of how unique they are. A relevancy of 0.0 would disregard total frequency, and only look at the most unique words to each topic. In this scenario, I found a relevancy of 0.1 to be the most interesting, but I would encourage you to explore on your own!
Once you have explored this visual, check out the individual Topic pages to see how each topic stacks up against the rest, and see where I have identified features of AirBNB listings that most uniquely represent each topic.