Amazon Launches Full Text Book Search

m00nun1t writes "Amazon have launched a new service that allows you to search the full text of books. This sounds like an incredibly useful function as well as technically impressive at this scale. I wonder if a patent is in the works." Or if a patent is already owned.

Comments

Fine grain searches take the adventure away

[Dancin_Santa]

Back in the early days of the web, when Yahoo was still a catalog of links and not some super news/search/auction/ebusiness/do-it-all website that it is now, searches were much more fun.

You really never knew what would turn up as you traversed the Yahoo directory structure. You start searching for blues music and you'd end up with a list of 15 or so good links with .wav samples and more than likely an artist you'd never heard of before. That was the best part, getting introduced to things you hadn't even thought to look for.

As search techniques are becoming more refined, we are now able to do specific word searches on websites and now books. That's fine if you know exactly what you are looking for. For example if you want to get that book about 'replicants' you'll find Blade Runner, but you won't find anything else. You won't get any information except exactly the thing you are looking for.

And I think that that is where the problem with this kind of search lies for books/music/etc. If you want to find a song or a book, it most likely isn't going to be a specific word you remember, it will be the tune or the plot, both of which are not searchable.

I don't see this improvement in Amazon's search system as that much of an improvement. A better improvement could be made to the 'We thought you'd like' feature. Instead of finding only what I'm looking for, I'd like to find other things I might also be interested in.

Potential tool for discovering plagiarism?

[Anonymous Coward]

I remember a teacher once telling a class I was in that our essays may be compared to other essays published online to check for plagiarism.

Granted, Amazon.com's feature will only (for now) include 150,000 books, but this may very well be another way to catch plagiarizers. Just type in a suspicious phrase and see if there are any 'hits'.

No more out-of-print books

[Bushcat]

As the digital index builds up, we will rapidly come across the situation where the electronic book is searchable, but the printed form is out of print. If this service ultimately allows single copies to be printed for delivery, it will be an outstanding demonstration of print-on-demand technology as advocated by the Print On Demand Initiative and others.

I'd love to be able to browse a giant back catalog, knowing that an original or facsimile copy could definitely be delivered to me.

Anyone else notice this?

[mike_lynn]

You have to have an account to view the pages. Fine, great. But then it brought up this screen:

By publishers' agreement, we are pleased to offer Amazon.com customers with a valid credit card the ability to view copyrighted pages.
Your account will not be charged.
This one-time process enables you to view limited copyrighted material through our Search Inside the Book feature.

So they'll let you browse the search pages, if you can prove your identity on record and provide them with financial information. No thanks.

Re:abuse

[wfberg]

How easy can this service be abused, with automatic webbots doing the searching?

Not so easily. It's easy to see why. The books will be scanned in using OCR. These days a fast and convenient and almost error-free process. But not entirely error-free. Good enough to find documents that are highly relevant to a particular keyword (if "hydraulics" occurs 9 times, what are the odds of OCR getting it wrong all 9 times?) but not good enough for entirely automated book-to-text.

If amazon would display highlighted portions of the books contents if would probably not exceed a few lines, just like google doesn't present entire webpages in it's result screen). If they did want to show more, they'd have to show an image of the scanned in page anyway, since OCR errors would not be very pretty. (A lot of digital archiving products use a similar approach; they index PDF files that contain the OCR'ed text, invisible to the end-user, and the scanned pages as content which the end-user looks at).

Besides, to search for each page of a book, you'd have to search for a keyword on each page of that book. Such keywords would most easily be extracted by scanning in the book via OCR anyway!