Having struggled in the past with LONG essays and meeting word counts, we decided to set out and find our own solution. That’s how the idea for Compressr was born.
What it does
Compressr condenses your input down to its most important sentences and gets rid of the irrelevant ones. It also gives you a glossary of the most important terms in the text with a hyperlink attached to their Wikipedia pages. You could use Compressr to optimize your cover letters and resumes to make sure potential employers only see what’s necessary, shorten essays to meet word counts, summarize research papers with a list of terms, and so much more. Compressr highlights what's important and strikes off what's not.
How it works
We have both a React app and a Node API running parallel. The Node API acts as an intermediary/interface between Google Cloud Services and our React app. We make a GET request to the Node API from React which itself makes a GET request to Google Cloud Services NLP API and consume the data on the node server which cleanses and processes the data through a custom algorithm we developed to rank sentences based on their importance and pull out important words. This cleansed data is then forwarded back down the chain to our React app which then highlights and strikes out sentences based on their 'rank' attribute. The results are displayed on the screen.
Challenges we ran into
We found that integrating all the individual elements into one package was challenging given all the different technologies we were making use of. Also, we didn’t have too much experience with CSS so front-end development was a learning and discover process.
Accomplishments that we're proud of
We are proud of completing this project in the given time frame and having incorporated the different elements/code into the final product. Since some languages were not entirely familiar for us, this process of learning and understanding was very rewarding.
What we learned
What's next for Compressr
We understand that not all text follows the same structure and the patterns vary depending on the medium of the work (eg. novels and poems). For this reason, we ultimately intend to fine-tune our methodology to adapt to these differences, by improving our algorithm to apply a more relevant ranking system to determine the value of each sentence.