Topic Modeling Streamlit App Deployed Via Streamlit Sharing (My First!)

Hi all,

Wanted to thank the team for providing a solution to what I’ve always thought was obvious target / gap in the current ML landscape. And for free! It’s certainly a solution I’d advocate strongly for internally at my company when Forgive my blog post if it seems a bit gushy but it’s really the solution I think is needed for a proof-of-concept stage type ML project even if immediate mode mechanics feel awkward for folks with prior GUI experience.

For anyone that is interested in NLP, topic modeling, Streamlit, or Streamlit Sharing I recently shared a blog post detailing my journey from idea to sharable web application with the aforementioned. A lot of fun and I’d appreciate anyone here taking a look and providing feedback.

Biggest issue I ran into was the lack of a clear idiomatic solution for session state (and I am well aware that there are out of core solutions / work arounds). Support for multi-page would be a number 2 since it might help on a regulatory front (e.g., I’m deploying an update to an app vice I’m deploying another app). I feel like a lot of this is complaining because of how fantastic the experience is already (I’ve spent a ton of time in typical callback style GUI frameworks).

Here are the links

blog post: Streamlit Topic Modeling | Dr. Bryan Patrick Wood's Website
github: GitHub - bpw1621/streamlit-topic-modeling: Topic modeling streamlit app.
Streamlit Sharing app:
LinkeIn: Create & Deploy a Topic Modeling Application Fast with Streamlit


Hi @bpw1621, welcome to the Streamlit community!

Both of these features are on our 2021 roadmap, with state coming first (I believe). Stay tuned!


@randyzwitch I was aware and that’s great news. However, the state problem I was experiencing I think is actually well addressed by the new submit button and forms feature that was just announced. Brilliant!

Lesson relearned: always ask what problem is trying to be solved vice what features should be implemented.

Keep crushing it: strong advocate here for the niche Streamlit very nicely fills.