Book Dragon Known Issues

Book Dragon Archive uses several systems to identify books, find information about them, and automate as much of the scanning process as possible. Those systems do not always agree with one another, and some of the information we depend on comes from outside services we do not control. Most books will scan and identify without trouble, but there are a few places where the limits of the available data, the physical book, or the current scanning system can affect the result.

Book Data Providers

Book Dragon checks multiple free and open data sources instead of relying on a single book database because no provider has everything. Using several gives us a better chance of finding the correct book and filling in information another source may be missing, but those providers are outside our control. A service may respond slowly, return incomplete information, or occasionally be unavailable, which can make a lookup take longer than expected or return less information than it normally would. If a lookup appears incomplete, trying it again later may produce a different result.

Series and Universe Information

Series and universe information is one of the weaker areas in the book data currently available to us. Providers organize books differently, series information is often incomplete, and larger fictional universes are frequently missing altogether, so two databases can contain the same book and still disagree about where it belongs. Adding more sources does not solve much when the sources contain the same gaps or organize the information differently, which is why Archive Nine is developing its own book information API to build a more consistent relationship between books, editions, series, and universes. This will be a gradual process because the data has to be collected, connected, and cleaned rather than simply copied from another source, but it should improve these relationships over time.

Genres

Genre data is one of the messiest parts of the public book information we currently receive. A single book may come back with ten or more genre labels, and those labels can include a broad genre, several subgenres, increasingly specific subgenres several levels deep, and sometimes categories that another provider would describe as a subject, theme, or audience classification instead. Different sources may also use different names for the same category or disagree about where a genre belongs in the hierarchy, so the result can be technically descriptive while still being difficult to use for organizing a personal library.

Clean, normalized book data is available commercially, but high-quality datasets cost money, and Book Dragon Archive is being designed around the goal of keeping its primary functionality free forever. Replacing the public sources with an expensive commercial service would introduce the kind of continuing operating cost we have tried to avoid elsewhere in the application. Our longer-term answer is the API system we are already developing for series and universe information, which will eventually give us a way to normalize genre names, understand the relationship between broad genres and their subgenres, and reduce some of the duplication and excessive specificity coming from outside sources. That work will take time because cleaning genre information is considerably more complicated than choosing one label from a list, so genre data may remain excessive, inconsistent, or more specific than expected while that system develops.

Automatic Page Capture

When you scan a page, Book Dragon looks for the edges of the page and waits until it has an image it believes is usable before capturing it automatically. Lighting, shadows, page curvature, the angle of the book, the background behind it, and how much of the page is visible can all affect detection, so sometimes finding that usable image takes a little movement. Repositioning the book or camera slightly is often enough to give the scanner what it needs, and manual capture remains available when automatic capture cannot get a reliable lock on the page.

Shared Barcodes

Barcodes are usually one of the most reliable ways to identify a book, but there are exceptions created by the way books have been published and cataloged over time. Some books share a barcode with another book, edition, collection, or product, while in other cases an outside database may associate a barcode with a specific ISBN even though the physical book being scanned is different. If the barcode itself points to that ISBN, Book Dragon has no additional information from the barcode that tells it the relationship is wrong, which means a scan can work exactly as intended and still identify the wrong book. When this happens, scanning the ISBN printed inside the book, particularly on the copyright page, will usually provide a better identifier for the specific edition in your hands.

Front Cover and Title Page Recognition

Front covers and title pages remain difficult to scan reliably because books were designed for people to read, not for software to identify. Covers may use custom lettering, decorative type, unusual spacing, words broken across artwork, vertical text, partial letters, or title treatments that depend on the image around them to make sense, while title pages can still use unconventional layouts and typography despite generally being cleaner. Book Dragon can use these pages as another path toward identifying a book, but they are currently less reliable than a barcode, ISBN, or copyright page. We are continuing to develop this part of the scanner, with larger improvements to front-cover and title-page recognition planned for version 2.