Do you have too much to read?
Do you hate when you invest time in a text only to find out it didn’t solve your problem?
AI generated summaries can give you a map of the text.
Imagine a non-fiction book.
It’s 500 pages but the title is intriguing.
You want to know if it’s worth your time before diving in. There are book reviews and summarization services like Blinkist or Insight App.
But what if you want to go deeper and see the original text in some parts?
What if you want to pick and choose what you skim vs what you read in depth?
With the AI summaries at various levels of the text, you can choose.
We have too much information to consume and too little time.
That’s where a tree-view hierarchical summarizer can help.
(the name sucks, I don’t know what to call this)
You can use LLMs to split the text into various chunks - by section, chapter, even paragraph. At each level the LLM will summarize, extract key points, and highlight new definitions.
Then you, the reader, determine if it’s worth diving deeper.
I like this idea but the UX needs work.
I’m not a front end guy but I can see this style of reading becoming a new way of working.
The summaries at each level require using various AI techniques like RAG and hierarchical summarization. The data processing can be abstracted away by an application.
The tricky part would be generalizing the solution to various types of text - knowing how to chunk the text properly.
What do you think?
Would you want to read this way when learning?
Do you know if something like this exist already?
If you’re interested in more, I have a 4 minute walkthrough of how I did this on YouTube.
My Attempt at a Tree-View Hierarchical Summarizer to Read with AI