N.B. Demo is in Google Drive link
The pandemic has hit the book industry hard and exacerbated existing systemic challenges for underrepresented authors. Without book tours, in-person readings, signings and festivals, for example, it's hard to get attention not only for book releases but also for new authors.
New authors already work hard to find an audience for their books, and due to the pandemic they now need to work even harder. While online book sales have gone up during lockdown, a number of those sales were facilitated by Amazon, where competition is sky-high and authors are charged fees. Even if their book is worthy of being a new york times bestseller, it may not have a big enough readership or enough reviews and publicity to rise to the top of searches.
To help promote books by underrepresented authors that are highly rated and have fewer reviews, we created Bookreads Plus
What it does
Bookreads Plus is an easy add-on chrome extension that customizes users’ browsing experience by helping underrepresented authors appear in searches.
Once the extension is installed, users can type in some keywords to search for books of interest and view a more balanced list of book suggestions. We were very intentional about not designing an algorithm to exclude any one category of authors. Instead, our goal was to analyze the initial results returned, evaluate books that received high ratings but fewer reviews, and add those results to the final results in an inversely proportional way.
How we built it
We used React JS for the front end and CSS for formatting. The search leverages Google Books API. In analyzing the Goodreads Kaggle dataset with the information about book ratings and number of book reviews, we also ran in GCP’s AI notebook.
Challenges I ran into
We had trouble importing Kaggle dataset into BigQuery, likely due to a need to clean the dataset. One team member fell sick as well.
Accomplishments that I'm proud of
For one team member, this was a first foray into performing searches via APIs and working with React! Coming up with a search algorithm that provides balanced results was also quite a challenge to tackle!
What I learned
In the discussion on ethics, computing and AI can be very powerful. When wrongly built, however, they can have major and lasting adverse impacts that leave us worse than when we started. We had a very hearty discussion of how to define underrepresented authors and how to go about obtaining sources and ways to objectively identify authors as being underrepresented.
What's next for Bookreads Plus
If we had more time, we would have liked to have connected our current search to a back end database to generate better book suggestions.