Glean

• 

  Glean

Created for Hack Day II by Alex Beals '18, Manmeet Gujral '18, Andrew Ogren '18, and Mike Ohene-Adjei '18

Motivation

One of the most pertinent issues we face in the current political climate is the rise of "fake news" and extremely polarized news sources that have come with the mass adoption of news distribution through the internet. While increased use of the internet as a news source has provided many with a more convenient, engaging, and personalized experience in news consumption, it has also produced the negative externality of "news" sites that disseminate biased misinformation. With this in mind, we set out to create an un-intrusive tool to allow users to evaluate the news article they are reading at a high level to understand the key points made, main entities referenced, and tone used in the article.

Function and Use

Glean is a Google Chrome extension. When the user visits a page, Glean allows them to view a 2-5 sentence summary of the article and general sentiment and entity sentiment analysis on the article to show the user what entities (e.g. Justin Trudeau, North Korea, NIH) are most relevant to the article and the sentiments (e.g. Anger, Disgust, etc.) attached to the entities. The user will be able to access this information by clicking the Glean extension in the Chrome toolbar and

Implementation

We chose to use a Google Chrome extension for two main reasons: first, Google Chrome is the most widely used web browser in the world and is very simple for development. Second, using a Chrome extension for Glean creates an unintrusive user experience that allows the user to use Glean without disrupting the web browsing experience. We believe Chrome's ubiquity is crucial to user adoption and general use and that Glean's unintrusive usability is key to user retention and encouraging use of Glean on a regular basis by users.

For the summary feature we used PyTeaser.py (https://github.com/xiaoxu193/PyTeaser) and the metadata description of the webpages to provide a clean and concise summary of the page.

For the sentiment and entity analysis we used Natural Language Understanding by IBM Watson calls on the backend of our project to analyze webpages.

Backend

The backend is ran using AWS Lambda to leverage the power of a serverless architecture, allowing easy scaling for the future.

Future Plans

After the hackathon we hope to refine Glean and take care of the UI/UX as well as edge cases involving websites that require subscriptions and unconventional website formats. We would also like to either create an app or a mobile interface as many people read their news on their mobile phones. Finally, down the line we would like to add more features like creating our own political spectrum analysis by aggregating data from users to learn what websites usually have what sentiment and political leaning using machine learning and existing data. Ultimately, we are fairly pleased with Glean and the idea and are excited to move forward with the project after Hack Day II.