Excited to share another milestone in our journey! This week, I developed a Streamlit app to integrate the ETL process and display sentiment analysis results in a user-friendly interface. šŸŽØ The UI effectively showcases the workflow, and the results are promising! šŸŸ¢


šŸ”§ This Week's Highlights:

āœ… Built and deployed the Streamlit app to visualize YouTube comment sentiments.

āœ… Streamlined the workflow for seamless integration of the ETL process.


šŸ“Œ Future Work:

šŸ”¹ Enhance the sentiment analysis pipeline to trigger the DAG automatically for specific users.

šŸ”¹ Deploy it on Streamlit cloud for the audience to play with it.


Check out the attached workflow image for a glimpse into the process! Iā€™ve also included some sentiment analysis results for the top 30 comments posted today on the "Jurassic World Rebirth | Official Trailer".


Looking forward to your thoughts and suggestions! šŸš€








Team Members:

Shaheer Airaj Ahmed (SDS Project Mentor), Jothi Mani Thondiraj (Project Lead), Vijai Murugan M, Patrick Edosoma, Funmi Ososanya, Oluwatunmise Olaoluwa, Youssef Benihoud, Valeria Salley-Washington, Jana ČadĆ­kovĆ”, Aryan Amin, Gabriela GutiĆ©rrez OlguĆ­n, Teslim Adeyanju, Ravi Venkatesan, and more.


#DataScience #MachineLearning #SentimentAnalysis #Streamlit #ETL #YouTubeAnalytics #Airflow #NaturalLanguageProcessing #AI #DataVisualization #PythonProjects

Admin

As said on LinkedIn, awesome update and highlights, this project is really top quality! Canā€™t wait to see the deployment on Streamlit šŸš€šŸ”„

Thank you Hadelin de Ponteves !

Thank you so much Edward Alvarez. Of course, I faced some challenges during the project, but I put my best effort into resolving them within the time frame.

Here are some of the challenges in this project:

1. Setting up the proper environment was initially challenging. However, with practice, I was able to manage it efficiently. On Windows, I observed that some users struggled significantly and needed more time to resolve setup issues.

2. Processing a large number of comments caused the response time to slow down. To address this, I limited the number of comments to 30, ensuring a faster and more efficient response.

3. The Huggingface RoBERTa model I used has a limitation on the number of tokens it can process. As a workaround, I had to filter out comments that were too long to fit within the model's token limit.

Please let me know if you have more questions.

Nice progress is evident here. Congrats. What (if any) challenges did you face in putting together this analyzer? Keep up the solid work. šŸ‘

Post a comment