Medium Profile Analyzer for writers using Streamlit and Python
Streamlit-based analytical tool for deep diving into your Medium Profile analytics
I have been writing on Medium for ~3 years now with around 80 blogs around AI & Data Science. As Medium doesn’t pay Indian writers, I never felt Medium Premium to be a dire requirement. Though, I always wished to have a tool that can help me get a deeper dive into my posts than what Medium provides in its free subscription.
What’s Next?
Using Streamlit, I recently tried designing such an analytical tool that can help tech & non-tech guys to have intermediate stats alongside some inferences to draw out.
Streamlit WebApp: https://mehulgupta2016154-medium-profile-analyzer-about-4kgv5r.streamlitapp.com/
Keeping things simple for now, The app has been divided into 2 parts
- Overall analytics
- Single blog analytics
Overall Analytics
The aim of this section is to give aggregated metrics. Below are the sections introduced for now
Blogs timeline
Given the year of release, this feature displays a timeline for all the blogs released in that year with cover photos & URL
Top Blogs
Given a criterion (view, reads, reading length, etc), which blog has the highest value
Blog %
This section is to understand what & of blogs fulfill a given condition (like 1k+ views, 1k+ reads, 10+ upvotes, etc)
Blog Frequency line plots for analyzing your writing pace across years/months
Word Cloud over blog titles only to get to know what you write most about
Correlation heatmap to understand what factors are affecting your growth/decline. For example, My read % was lower for blogs where the length of the blogs was high. It may be indicative of the fact that people don’t prefer reading long texts
Bi Variate Clustering for important factors. You can add other combinations as well
Distribution of different factors using boxplots
Single blog Analysis
This section focuses more on a particular blog which includes
- Word-cloud based on blog content for the chosen blog
- Brief stats around total views, reads, upvotes, etc
- Keywords using Yake (unsupervised statistical model for keyword extraction from text)
Blog comparison
This sub-section includes a one-on-one comparison between 2 blogs based on factors like total views, total reads, etc. This section also includes a similarity score between the two blogs using
- Sentence BERT for generating sentence embeddings
- Cosine similarity
Many of you may face some glitches in the app as it hasn’t been tested on many users. Kindly report any bug on this post or on the git repo. Will be adding some major features in the coming time. Stay tuned !!
By the time, do read out:
- Building portfolio using Streamlit
- Graph Analytics (6 parts)
- NLP Algorithms(9 parts)
- Starting off with Time-Series (9 parts)
- Reinforcement Learning Basics (6 parts)
- Generative modeling (7 parts)
- Transformers in NLP(4 parts)
- Dimension Reduction (3 parts)
- Object Detection using YOLO (3 parts)
- Tensorflow for beginners (concepts + Examples) (4 parts)
- Statistics for beginners (4 parts)