In this project we try to answer the question of "What do people think of the Seattle Public Library? If we want to explore what do the people think, the first and naive thing that comes to mind is visiting the yelp website and reading the reviews. The more efficient solution is to compress the reviews into one visualization. The visualization will have text from the reviews and images the users took and uploaded themselves.

The first step to tackle this problem was to download the reviews through the Yelp API. Unfortunately, upon checking the Yelp API, I found out that the Yelp API does not provide access to the full review text at this point. At that point, I decided to build my own database. The database consisted of three columns: Review (contains the review text), Images (contains the image names associated with that review), and Image Tags (contains the user-generated labels associated with the images).

The second step was to extract review tags from the review text. Since most of the review texts contained a lot of unimportant and common words, I used manual feature selection to get the most important features of the review. The third step was visualizing the review features and the images uploaded by the users.

Some of the reasons why people like the SPL:

• amazing architecture
• 10 floors of pure awesomeness
• 10th floor highest view point
• 9th floor map room
• really short bathroom stall doors
• worth the stop
• Beautiful library
• plenty of seating
• meetings at the auditorium
• great resources for business and market research
• helpful librarians
• click to chat feature on website
• good customer experience
• Seahawks themed library cards.
• best library
• crazy fast wi

The idea in the visualization is positioning the extracted features and the images around the Seattle Space Needle which is an observation tower in Seattle, Washington, a landmark of the Pacific Northwest, and an icon of Seattle.

I positioned the title "Why do people like the Seattle Public Library?" in the middle of the visualization. The images were placed above the title. Underneath the title, I positioned the extracted features from the review tags using the library word crammer. The library positions the text and also controls the size of the text by positively correlating it with the number of repetitions in the text corpse. When the visualization process is run, the words start to appear incrementally as seen in the first image in this section. Pressing any number in the range 0-9 will result in changing the images at the top of the visualization so that you can view different images from different users.

All work is developed within Processing
Source Code + Data