Ever get tired of spending hours staring at your mobile screen every day reading news articles? And to top it off, many news websites will spew lengthy articles beating about the bush, before getting to the actual meat of the content. For example, you'll end up reading a 100 lines just to know what that new Android or iOS feature is about. I, for one, am definitely tired of this loquacious attitude in news articles.
What it does
Saber is a web browser that summarizes the URL that you provide to it and will read them out aloud. You can always have Saber read out the full content of the article, in case you feel that you need more context about the content.
How I built it
- I used the dev-alpha version of the Flutter library to build a Linux desktop, Android mobile, and web app.
- In order to extract the webpage contents, I re-used the concept from my code that I had earlier written in Python: Mycroft AI Webpage Summarization Service. NOTE: I have only re-used the same concept. I have not directly used any code from my previous work. All Dart code submitted was written during ShellHack 2020's hackathon period.
- It uses Google's PEGASUS Model to perform abstractive text summarization on the content of the given URL.
Challenges I ran into
- I did not know Dart or Flutter before this week.
- I did not know much about UX design principals.
- I had never used Google Cloud APIs earlier.
What I learned
- After attending the Flutter workshop, I learned the Dart programming language over the weekend.
- After attending the UX / UI workshop, I learned about how to make interfaces that are empathetic to the end-users.
Accomplishments that I'm proud of
This hackathon inspired me to learn a new programming language and framework in a few hours over the weekend. Now I know that I can learn many more things if I put more focus and direction into the things that I do.
What's next for Saber
- I would like to add more focus on data privacy.
- So, I would anonymize the data sent to Google Cloud so that customers cannot be identified based on their browsing history.
- In the same vein, I would like to add a local TensorFlow RT model that can perform text summarization locally on the user's machine, in case the user doesn't want to use a cloud service.