Reading the news can be a daunting task. Reputable news sources often produce lengthy and tedious articles, discouraging children from gathering a global perspective. We set out to create a means for children to stay informed and engaged with the state of the world in a more friendly way.
What it does
Newsreel summarizes news articles into six panel comics detailing the most important aspects of the story. Our Chrome Extension allows users convert any article they come across to a simple comic strip. A user can visit the Newsreel website, enter the URL of an article, and read it as a short comic. If the user inputs multiple articles into Newsreel, he can flip through them as a bundled comic book. Website users can also select one of our fifteen daily digests of popular news sources. Users have the option to save Newsreel's generated comic strips for later, allowing them to share them with friends or read later! Finally, if users would like to read more about a story, they are able to access the full article of any generated newsreel at the click of a button.
How we built it
We set up two servers on
glitch: one to set up and search articles and another to generate images for the comic strips. We created a static website and chrome extension on top of of these services to allow users to access comic versions of their news in different ways. We wrote a script to pull daily popular stories from a variety of news sources and present each in a Daily Digest comic form. We attempted to improve on the summary and image generation by using data-mining practices and set up a testing suite to compare our custom modules with the existing libraries.
Challenges we ran into
We faced challenges with lengthy processes - from our MVP, significant optimization was required to generate accurate comic strips at a reasonable rate for the user. Image processing initially took over fifteen seconds, and each of our summarizing algorithms took nearly ten seconds to run.
Accomplishments that we're proud of
Implementing two of our own algorithms to summarize articles in six sentences!
Optimizing image search and reducing image load time by a factor of five!
What we learned
Ayan learned about
Jaccard similarity algorithms!
Vinesh learned how to implement
k-means clustering in
Haosheng learned about image processing and how to speed up Google queries!
What's next for newsreel
We'd like to improve our image identification algorithm and further optimize for faster loads. We would also like to implement comic series, where we follow a single event over time and update a larger comic book to display real time changes.