Summarizer NLP

• 

  Once the user uploads an image, we extract the text using Tessaract OCR, and then feed it into our co:here NLP model to get the summary.

• 

  Once the user uploads an image, we extract the text using Tessaract OCR, and then feed it into our co:here NLP model to get the summary.

Inspiration

Long documents are daunting. Whether they are textbook chapters, PDFs, or lecture transcripts, digesting the information can be a tedious task. We created the Summarizer to solve this!

What it does

Studying for a test? Assigned to read a 30 page paper? Take a picture of the pages or upload the article to reduce all the information down to a fraction of the size!

How we built it

• We used the "co:here NLP API" to easily train models, interpret text, and summarize the text
• Tessaract to ocr text, converting text in the image to plain text
• Flask and Python 3 for backend and website
• CSS & HTML for styling

Challenges we ran into

• OCR was spotty sometimes depending on image quality
• Model training was difficult for different subjects

Accomplishments that we're proud of

• Created a working product.
• Added a front end to working back end

What we learned

• NLP and drawing conclusions solely from text input is really cool!
• How to get an quick MVP up and running
• Flask is a really smooth framework

What's next for Summarizer NLP.

• Implement as Chrome extension for easy seamless summarizations
• Instead of summarizations, use NLP to translate text and place directly back onto the image.
• Integrate summarizations into input to create new textbook/articles with all other original content (pictures, titles, citations, etc)
• Backend file parsing to accommodate 2048 token limit
• Paywall/pricing model

Built With

• cohere
• flask
• natural-language-processing
• python
• tessaract-ocr