Condensify

• 

  You can send the document for signatures, and it accommodates multiple signers.

• 

  Files can be in any format. Signing is completed without needing any installs or even having to log in.

• 

  It offers tracking across multiple stages, enabling you to monitor updates and synchronize them with the dashboard for easy management.

• 

  Comprehend documents effortlessly without the need to read every line and extract intricate details.

• 

  To initiate the process, simply drag and drop your files, whether they are images or PDFs.

• 

  Upload as many documents as you like with customized context and chatbot for each.

• 

  Files that are uploaded undergo OCR scanning, and the app generates document summaries.

Elevate your document comprehension with Condensify. It harnesses the power of generative AI to help you grasp complex materials. It learns from your uploaded documents to create a customized chatbot which can answer questions about the document and making you more confident when dealing with important paperwork.

Inspiration

All grown-ups find themselves faced with a heap of paperwork at some point. It's nearly impossible for any of us to thoroughly read and fully grasp the legal ins and outs of every document we're asked to sign. Plus, the time crunch often leaves us with minimal opportunity to do our due diligence. Some fortunate folks and organizations can bring in experts to handle this, but for the vast majority, that's not an option. This is where LLMs can come in to lend a hand and make this whole process smoother for everyone.

What it does

Our system scans documents for you, extracts the relevant information, and uses it to create a context that makes it much simpler for people to understand extensive, wordy documents that carry significant real-world consequences.

How we built it

We take the files you upload and use OCR technology to extract text from them. This means we can scan not just PDFs but even handwritten documents of reasonable quality. We then provide this scanned text to our LLM to establish the context. Using this context, we generate a summary and set up an interactive session that allows you to dive into the document and learn from it as you explore. We utilize Google's powerful generative text APIs, to help with the process.

Challenges we ran into

LLMs bring a degree of unpredictability to the table, and this can pose certain challenges. Additionally, our API currently has size limitations that prevent us from effectively handling very large documents, even though such situations are where a tool like this would offer the most significant value.

Accomplishments that we're proud of

We haven't experienced many instances of "LLM hallucinations" up to this point, and this could be attributed to the solid grounding provided by Google's API.

What's next for Condensify

I encountered the premise of this solution as a personal challenge, but the proof of concept appears to hold significant promise. Further development and exploration in real-life scenarios seems possible. The API limitations are one challenge, but maybe also more prompt engineering may help. Document updates and follow-up signing can be integrated as well.

Files used for testing: https://github.com/O-LAP/reward-wave/tree/master/public/test-file

Built With

• bard
• firebase
• javascript
• node.js
• vuejs
• web