In today’s world, there is a constant need to keep up with the latest happenings around the world to stay informed. Doing this is becoming tougher and tougher in the ever-expanding amount of information. This begs for the need for filtering such information, so as to make this quest less scary. Also, being able to test the comprehension of new information would go a long way for actually learning new things.
What it does
It tries to formulate questions which can assist users in finding the information that they are most interested in. Basically, lets users choose the articles via questions that intrigue them. This particular chrome extension is tailor-made for hacker news (https://news.ycombinator.com/). The web site hosts a lot of valuable information but is not formatted in the most conducive way. Through this chrome extension, we go through all the articles and display interesting questions about the article right under the link to that article. We believe this makes it easy for the users to decide on which article to read. The extension also features a playground where any article could be put to make interesting questions out of it. This can also be used to study for subjects like history or a driver’s license exam where the most important thing is to remember facts could be learned.
How we built it
The chrome extension parses the hacker news website and sends back URLs of all the articles that are there to an API hosted on google cloud platform/Heroku. The API backend then gets all the content of the URLs and summarizes it using an API from deepai.org. This summarized text is then put into the Google cloud natural language entity API, which gives all the relevant entities. Then with the help of our algorithm, the API makes appropriate questions to return to the extension. The extension then adds a new child below every article title which has questions.
Challenges we ran into
Lack of adequate research in the domain of making questions from sentences/articles.
Accomplishments that we're proud of
We were able to successfully build the 1st version and host it on Heroku and Google Cloud Platform. Quality of the questions turned out to be acceptable. We were able to successfully implement what we wanted on a live website. (hacker news).
What we learned
We had an extremely positive learning curve while building Interrobang. NLP, one of the trending topics recently, was fun to work on. Not just technology wise but we got an amazing exposure to enhance our entrepreneurial skills. The key takeaway from this project was to spend quality time on the problem itself, understanding it and iterating until it is clear instead of coding right from the first minute.
What's next for Interrobang
Interrobang would then be launched with Node instead of Flask for better performance. We would like to constantly improve the quality of questions using "ab" tests and provide support for a wide variety of browsers.