Original article was published on Artificial Intelligence on Medium
The Solution — Curation Engine That ‘Helps’ 😎
For an advanced platform like Medium, having to actually read through the Medium curation guidelines followed by additional guidelines from individual publications is sluggish and cumbersome. What I am proposing is to create an Artificial Intelligence (A.I) powered Curation engine that automates the majority of this process and works for the Writers.
Those amongst us who have been writing for a while on Medium would have most certainly heard and used Grammarly. Grammarly integrates seamlessly with the Medium platform and has become an arsenal in the hands of every seasoned writer. Grammarly’s advanced systems combining rules, patterns, and artificial intelligence techniques like deep learning, and Natural Language Processing (NLP) have become a tool loved by writers.
For those who are new to the concept of Natural language processing, it is a branch of Artificial Intelligence that is all about teaching machines to understand and process human language. Various applications such as machine translation, sentiment analysis, essay scoring, etc. use NLP.
The Solution — In Images
Below is a basic visual representation of what the system could look like or should aim to achieve in my opinion:
Step 1: Right before you publish an article, the system provides an interface that scans the article and compares it against the Medium Curation guidelines and other parameters:
Step 2: Based on the outcome, it either presents a positive response or suggests areas for improving your chances of being selected for distribution in topics:
High-Level Steps in Creating the Curation Engine
- Step 1: Activate the Neural Network (Curation Engine) for training.
- Step 2: Divide the Input data set (previously curated and non-curated articles on Medium) into 70% for training and 30% for testing.
- Step 3: Input the training data along with Medium Curation Guidelines to the Curation Engine.
- Step 4: Curation Engine analyzes and learns based on writing styles, previous decisions, and tags. Tags will also be an important input in the learning process (For example, an article under ‘Love’ tag would have a very different feel to an article under tag ‘Technology’).
- Step 5: Once the model is trained, the accuracy of the system is determined by running on test data.
- Step 6: Once optimal accuracy has been achieved, the Curation Engine is rolled-out on the Medium platform.
- Step 7: The Curation Engine starts providing personalized recommendations to the Writers to improve their chances of being distributed in topics by Curators.
- Step 8: Accuracy of the system is continually improved as more dataset is continually fed into the system.
With millions of subscribers and articles being published frequently, there is a lot of data being generated continuously. This can further refine the AI model to recognize patterns and learn collectively from all writers and individually from each writer’s style. This will allow the Curation Engine to provide personalized feedback and recommendations tailored to each Writer.
Just like Medium is able to provide a personalized experience to its readers, the Curation Engine can provide a personalized experience to its writers.
The Solution — Taking It Further
Often, Medium publications have additional guidelines that Writers must follow before submitting a draft for consideration.
Hence, in addition to comparing the article to Medium’s curation guidelines, The Curation Engine can be further developed to allow publishers to add their guidelines on top. This would make the curation engine further powerful and useful.