Kindle Review – Kindle Fire Review, Kindle 4 Review

This post isn’t talking about how to use the Internet to connect readers with authors. It’s about a more straightforward channel – Reader to Author via eReader. It’s the equivalent to your books or your home library coming alive and getting you in touch with the authors most appropriate to your tastes.

Finding authors or readers on the Internet involves too many extrinsic things

Let’s say you’re a reader looking for a great military Sci Fi book and that book happens to be Dominant Species by Michael E. Marks. You just don’t know it yet.

• A good book search system should be able to able to take you from ‘good military sci-fi’ to Dominant Species or another suitable book instantly - Without any detours or delays.
• A good recommendation system should suggest Dominant Species or another good book as soon as you finish your current book.

That makes sense because you want the shortest path between your current book (or your intent) and the next book you will read and really enjoy. In this case the journey to your next great read is not the destination.

The Internet is non-0ptimal for both search and recommendations

On the Internet there are a lot of diversions along the way -

1. The first diversion is having to put down your book and log on to the Internet.
2. The second  problem is choosing between the options presented in a search engine and then spending 15 minutes parsing forum threads and blog posts.
3. The third problem is availability and pricing – The book may or may not be available for your eReader or at your nearest store. There’s also a chance it’s priced outside your budget.
4. Perhaps the biggest problem is that there is no personalization to your tastes so you don’t know for sure that you will like it and you definitely don’t how much you will like it. 

When all is said and done – You might find a good read 40 to 50% of the time.

Recommendation systems on the Internet also suffer from lots of problems -

1. You have to seek them out i.e. they don’t come to you.   
2. They are either store based or social network based. That means they are limited in their knowledge of your tastes.  
3. They don’t dissect books – Most recommendation systems use surface level things like what other people with similar purchase history bought and your reviews of similar books. They don’t break down the elements of a book.  
4. The amount of feedback you can give is limited. 1 to 5 stars, and the fact that you actually bought the book - that’s it.  
5. They aren’t good in two specific areas – Finding you books outside what you normally read, Ranking the books that are recommended based on how much you might like each.  
6. They do not factor in other potential clues i.e. what movies you like, what newspapers and blogs you like, what your beliefs and interests are. 
7. They aren’t a complete solution. It’s ‘head west’ rather than ‘head west for 500 metres, take a slight left, keep walking for another 200 metres, and it’s on your right’.

Recommendation systems will probably suggest a good book often enough to be useful. However, they are flawed in several critical ways.

If you think of ‘finding the next book’ as a problem, search and recommendation systems are time consuming and somewhat inelegant solutions.

It’s just as difficult for authors to find readers

Take an author who writes a mystery novel and tries to find an audience for it online -

1. He has to put together a website, figure out how to rank in search, understand all this social media stuff, and generally become an expert in something completely outside of writing. 
2. Across the Internet (both in search and social media) there’s an ageing factor i.e. it takes a certain amount of time to gain trust – both from algorithms and people. 
3. He has to compete with a lot of sites and people who aren’t even selling books. Scammers, spammers, marketers, and lots of other people who want to get readers to play a game or buy a magic health potion instead.
4. There’s no way for him to identify his specific audience i.e. You find a forum and then have no idea who actually reads books because everyone seems to be promoting their own book.
5. He has to do all of this without raising any flags – too aggressive in a forum and readers get upset, too aggressive with search engines and the site gets penalized.

It takes people years to figure out and specialize in marketing and social media. An author is expected to do all this and also figure out reader psychology and write and polish a book.  

It’s an endeavour Jason and the Argonauts might balk at. 

The amazing thing is that readers are dying to find good books to read – At its heart the reader to author connection should be an exceedingly simple one to create. Yet it’s complicated because the Internet is overwhelming us with options and information and distractions and existing solutions are approaching things from a non-optimal angle.   

eReaders and an in-built intelligent heuristic might be the solution

With eReaders we get a device that is built for reading and a channel that is devoted to finding and buying books. It’s crucial that most distractions are ruled out. It’s also crucial (in the case of the Kindle) that it’s a closed eco-system that keeps out people of bad intent.

Not only do we have a perfect channel for connecting readers and authors we have a wealth of information on readers’ reading habits and purchase history.

The eReader is a direct channel and it knows exactly what you read 

By direct channel we mean that there could in theory be a direct connection between the reader and the author. The channel/platform is the enabler and since its’ interests are aligned with that of the reader (find a good buy, buy it, read it) there is minimal friction.

Equally as important is that the eReader has all the data it could need to make recommendations for you or to refine search results when you are searching for new books.

Here are the things that are available -

1. What you bought.  
2. What you opened and actually read.  
3. What books you finished, what you left halfway, and what you read multiple times.
4. How long you took to read a book after buying it.
5. What you rated the book and the words you used in the review (if you added one). 
6. How many books you have on your eReader. Which types of books you read the most.
7. Your favorite authors.

There’s actually a lot more information locked up in there – The data locked up in your eReader can almost certainly predict your tastes and your future purchases better than you can.  

Here are the additional things that could be collected and used -

1. At the end of the book your rating of the book and whether you thought it was good value for money. 
2. Tags for the book so you can associate the words and feelings you felt with the book. 
3. Your other purchases (from Amazon or other stores).
4. Your Internet browsing on the eReader’s browser and your bookmarks (imported from your computer’s browser).
5. Data from your social network profiles.
6. The option to add information on physical books you read.

Again, there is a lot of information that can be collected. Even if we stick to just the first 2 points there’s a wealth of information.

An Intelligent Heuristic that resides on the eReader

Once we have all that information we combine it with two things -

1. A categorization/tagging system that dissects a book and analyzes it and associates a hundred different tags with it that describe the book from a multitude of angles.
2. A heuristic that uses the tagging system and the users’ reading history and available books to recommend new books and refine book searches.

What’s the categorization/tagging system?

1. It takes a book and looks at it from every angle – the type of language, the style, the genre, the themes, the words used most often, the pace of the plot, the frequency of adjectives and nouns and verbs, the number of words used, the difficulty level, and a whole lot more.
2. It then tags the book with a hundred (or more) different tags that explain the structure and qualities of the book.
3. Perhaps it even adds on things like the feelings the book evokes in readers, the words the author associates with the book, and so forth.
4. At the end of the process we have a collection of words (tags) that best describe what the book is, what it does, and what people feel about it.

The tagging system gives us the book’s DNA.

What is the heuristic?

1. The heuristic says that if a particular book’s tags overlap in certain ways with the tags of the books that a user has really liked (the user’s favorites, the ones the user rated well, and so forth) then it’s very likely the user will love the book.   
2. There can also be negative tagging i.e. books with certain tags are never finished or they are bought and never read.
3. Some tags can be figured out by analyzing the book itself, some tags have to come from people who’ve read the book, and yet others come from various sources like the retailer, the author, and the publisher.
4. The heuristic is basically an evolutionary algorithm. It starts off from a common template and then it evolves according to each user’s needs. If a users cares a ton about intricate descriptions and detailed character development but doesn’t care at all about length of the novel or difficulty of the language the heuristic adapts to these preferences. Note that this information is gathered primarily from reading patterns and purchase behavior.

The aim would be that the heuristic keeps monitoring the user’s reading and in particular how the user reacts to the books recommended and adapts and molds itself to the users’ tastes.

Since it’s on the eReader itself there are no privacy concerns – The user could choose to keep his evolved heuristic private.

The Tagging, the evolutionary heuristic and residing on the eReader are the key parts

It’s a recommendation system (and a search refining system) that would really work. The 3 key parts are -

1. A very effective way to break up books and tag them. Get the book’s DNA.
2. An evolutionary heuristic that evolves to almost perfectly capture the user’s taste in books.
3. Residing on the eReader so that we’re in touch with all the user information and can provide direct recommendations (and direct refining of search results).

It’s a system that might take a few months (5 to 10 books bought and read) to learn a particular user’s tastes. However, once it has a reasonable amount of data it will blow away every other recommendation system. Plus it would keep getting better. You’d have a book’s DNA and you’d have a user’s tastes and you’d match them without any distractions or interference.

Filed under: evolution Tagged: | pandora for books