Building AbbrivioAI: Harnessing AI to Condense Textual Complexity into Clear, Actionable Summaries.

Hello, LinkedIn community! I'm thrilled to share my latest project, AbbrivioAI, a text summarization web application built using React and Django. This journey has been both challenging and rewarding, and I'd love to share some key highlights and learnings with you all.

Project Overview

Project Name: AbbrivioAI

Tagline: Harnessing AI to Condense Textual Complexity into Clear, Actionable Summaries

Frontend: React

Backend: Django

ML Models:

Encoder-Decoder with Self-Attention Mechanism

Bidirectional LSTM (from scratch)

Fine-Tuning: BART Model

Training Environment: Google Colab Pro

Challenges and Learnings

1. Building Models from Scratch

Encoder-Decoder with Self-Attention:

Significant effort was put into this model, but achieving satisfactory results was challenging due to computational limitations.

Bidirectional LSTM:

Faced hurdles with large datasets and extensive training epochs due to resource constraints.

2. Fine-Tuning Pre-Trained Models

BART Model:

Fine-tuning BART provided better results, but I was limited to a subset of the dataset due to computational restrictions.

3. Training Environment

Google Colab Pro:

Provided a good platform for training, but proved insufficient for extensive training requirements, prompting the need for more robust solutions.

Next Steps

Exploring Paperspace:

To address computational limitations, I plan to switch to Paperspace for more robust ML training capabilities.

Improving Accuracy:

Continuous efforts to enhance the model's accuracy through further fine-tuning and dataset expansion.

Experimenting with Pointer-Generator Method:

In my own model, I am planning to try the pointer-generator method and train for more epochs using the whole dataset.

Deployment:

Planning to deploy the application using AWS or Azure for better scalability and performance.

GitHub Repository

For those interested in exploring the technical details of this project, you can find the code on my GitHub repository: GitHub Repository Link

Key Takeaways

Resource Management:

Ensuring adequate computational resources is crucial for training large models effectively.

Model Selection:

Fine-tuning pre-trained models like BART can significantly enhance performance, especially when working with limited resources.

Continuous Learning:

The field of AI/ML is ever-evolving, and staying updated with the latest tools and techniques is essential for ongoing improvement.

Seeking Suggestions and Collaboration

I'm on the lookout for co-op opportunities in the Artificial Intelligence/Machine Learning field and would greatly appreciate any suggestions or feedback to help improve this application. If you have experience with similar projects or any insights on overcoming these challenges, I'd love to connect and discuss further.

Feel free to reach out with any queries or suggestions. Your feedback and support can help build a better application and advance my learning journey.

Thank you!

#AI #MachineLearning #TextSummarization #React #Django #GoogleColab #Paperspace #AWS #Azure #CoOp #TechCommunity #LearningJourney