Thursday, March 1, 2012

Flickr: BHL's Unexpected Success Story

In July, 2011, several members of the BHL staff began putting illustrations from BHL books on Flickr. It started as a simple way to flag favorite staff images and document the illustrations we were using in various outreach activities, like Twitter and Facebook. Today, our Flickr presence has grown to a collection of over 24,000 images and has become one of the most user-celebrated aspects of the BHL project, prompting such statements as "I think I'm in love" and "My favorite Flickr in the whole world!" Several people have asked us to share our experiences using Flickr as a component of BHL and discuss its growth from inception to the success it is now. We couldn't think of a better place to do this than on our blog.

What's In It for Us? The Benefits of Flickr

Using Flickr has generated a variety of benefits for the BHL project, many of which were unexpected. First and foremost, Flickr provides us with an easy way to showcase our amazing images - a BHL treasure that was all but hidden in our collection. Before Flickr, there was no simple way for users to access the images in BHL. Now, using Flickr sets and collections, users can go to one place and quickly browse thousands of images. And all of these images are available for free download and re-use, which is an excellent "selling" point for anyone interested in finding gorgeous natural history and botanical illustrations.*

Flickr also allows us to broaden our outreach potential, branching outside of our classic BHL interface and reaching our users on a new platform. The audience using Flickr is often quite different from our traditional user base, thus having a presence on Flickr means that we are able to reach a whole new group of people, including artists and k-12 educators.

Our Flickr group has also provided many unexpected benefits to BHL staff. Not only can users more easily find BHL images, but so can we! What started as a way to document those images we were already using in social media quickly became the source of much of this activity. Flickr images provide excellent fodder for our various social media activities, including tweets, facebook posts and trivia, and provide inspiration for many of our Book of the Week posts. Staff can now easily find illustrations to "spice up" BHL promotional materials, including brochures, business cards, banners, and merchandise, or even bring flair to events like donation campaigns. Plus, incorporating images into BHL reports transforms the otherwise boring white pages into alluring, eye-catching documents that people actually want to read.

The World of Tagging Images: Machine Tags

Through the use of machine tags, Flickr also allows us to collaborate more extensively with projects like EOL. Machine tags look like:
  • “taxonomy:binomial=Aegotheles savesi” (quotes required due to space b/w Genus & species)
  • taxonomy:family=Fabaceae
  • taxonomy:genus=Rhinoceros
  • taxonomy:common=Cardinal
Any time an image from our Flickr photostream is machine tagged, it becomes "flagged" for inclusion into the EOL Group Pool and will thus become a part of EOL species pages. (Images are added to the EOL pool on a monthly basis).

See this example of an EOL page for Corvus crassirostris. You'll notice that the image of the Thick-billed Raven from the BHL Flickr set Album of Abyssinian Birds and Mammals is featured on the page. How cool is that?! In addition to the links from EOL pages (under the "Literature" tab) to taxonomic name occurrences in BHL, now BHL image content can be repurposed directly onto EOL species pages.

We would absolutely love your help tagging more of our images for EOL. For more information about machine tagging, please see the instructions provided by EOL here and feel free to contact us via Flickr mail, Twitter @BioDivLibrary, or our feedback form if you have any questions. Please note that you have to have a Flickr profile to add tags to BHL images. If you are already a Flickr member or considering becoming one, and would like to have your own images of flora and fauna harvested to EOL species pages, please see the following for more information: http://www.flickr.com/groups/encyclopedia_of_life/rules/.

Proud Parents: Watching our Flickr Mature


Since the inception of our Flickr group, there have been many project milestones, the first of which (besides actually starting the account!) was implementing a basic organizational structure. Once it was clear that BHL wanted to fully participate in the Flickr environment, staff transformed a haphazard pool of images into organized sets and collections, with sets structured around individual titles and collections around subjects and content types (like "Mammals" or "Book of the Week"). Slightly later, staff capitalized on the opportunity Flickr offered us to celebrate specific events, organizing themed collections such as Halloween or the Holiday Season.

Pagination display on BHL
Eventually, staff realized that Flickr could help us target specific content in our collection for metadata improvement - specifically, pagination improvement. Pagination refers to the display of page numbering that you see on the left hand side of the BHL screen when viewing a book. Although very basic pagination is performed at the time of book scanning, BHL staff manually update this pagination with more helpful information such as identifying illustrations, designating articles, and embedding year and volume information. Learn more about BHL pagination in Gilbert Borrego's post. Since those books featured on Flickr became some of the most popular in BHL, and since it was difficult to locate illustrations within these items without improved pagination, staff began to prioritize Flickr titles for pagination improvement. As of this week, all 450+ titles in Flickr have been manually paginated in BHL!

Our most recent milestone addresses one of the most common requests we've had regarding Flickr content. Until recently, images in Flickr only linked back to the volume level in BHL. This meant that the BHL urls associated with an image in Flickr did not link to the specific page in BHL, but instead to the "item" or book that contained the image. It was often difficult to then locate the image within the book, especially if the pagination had not yet been manually updated. Thanks to the Technical Development team at the Missouri Botanical Garden, all images in Flickr now have page-level links associated with them, allowing users to find the page within a BHL book with a single click.

This improvement came with an added benefit for BHL staff. Previously, to upload images into Flickr, staff had to first download the images from BHL, open them in Adobe Bridge to embed metadata, and then bulk upload to Flickr (more on the previous Flickr process in Gilbert Borrego's post). Now, staff can flag illustrations directly within the BHL back-end administration portal. Software developed by the MOBOT team then automatically embeds book and page information to these images and uploads to Flickr. For more information on the technical aspects of these improvements, see our FAQ page.
__________________________________________

Overall, we've found that Flickr offers us an excellent opportunity to provide access to aspects of our content that were overlooked and under-appreciated. Reaching new audiences and demonstrating that BHL has much to offer outside the realm of taxonomy is a critical step in the growth of the project.


*All BHL images are public domain. However, Flickr does not provide a "public domain" license option. Therefore, our images are listed in Flickr as AttributionNoncommercialShare Alike Attribution, Noncommercial, Share Alike

13 comments:

CN said...

Terrific post on how the use of Flickr has progressed with BHL. Q: are species tags next? The plates have no names showing up, both on Flickr and on BHL (uBio). Where do you think tags should start: OCR on plates, or manually in Flickr--or from EOL? The crow you give as an example is named only at EOL--done manually? Could it be roundtripped back to BHL?

BHL said...

Great questions CN and thanks for taking an interest in our blog post.

Currently all species name tags are added manually to Flickr images. Automating species name machine tags would be nice but there are a variety of challenges to implementing this.

Often times, the species name is not available directly on the plate. OCR on plates is also difficult to achieve as the text could be rotated in different directions or perhaps in very small print, italicized, etc. making it difficult for the OCR to read correctly. In addition, there are a range of machine tags that can be used such as "taxonomy:binomial=Genus species", taxonomy:genus=Genus, "taxonomy:common=Common name phrase", or geo:lat=latitude which requires some form of interpretation before deciding which machine tag to use for which text string.

The are advantages to using the crowd to manually add machine tags as these tags are more carefully curated and more accurate than automated tags.

We hope to someday repurpose the tags that are in Flickr back into BHL and have submitted proposals to get support for this work in the future.

Chris Freeland said...

Hello, BHL Tech Director Chris Freeland here. BHL maintains bidirectional connections between its pages, the illustrations loaded to Flickr, and the images submitted to EOL, meaning that a piece of data added to one can be pushed to the others. I say "can" and not "are" because as the official BHL responder said, we have grants in consideration to make the actual data exchange happen at scale. In theory & structure it is possible; in practice we're not there yet.

So, the best place for now for tagging of scientific names & common names is via our illustrations on Flickr, and doing so using machine tags. Images machine tagged with a scientific name are "automagically" posted to the appropriate page in EOL. We are also working out how best to match up images tagged with a common name (which is often all that is present in the plate, or even in the list of figures that is often separate from the plates). Go ahead and machine tag common names because we'll be connecting those dots with the grants previously mentioned.

Chris

Jennifer W. Hanson said...

Thanks for this terrific post on the BHL Flickr experience. I've been on Flickr since 2006, a BHL fan ever since I heard about it, and a natural history and book enthusiast since childhood. The Flickr interface provides a great way to view the beautiful images in these old books, and I enjoyed reading this post about the ways the process works and is being tweaked behind the scenes.

I machine tag BHL images on Flickr when I have a little extra time to sit down with a set of images and work through them. One thing I've started wondering about is how the machine-tagging process works when there is a plate with lots of different species on it (an example is plates from Holland's The Moth Book). It's probably not useful for EOL purposes if a machine tag associated with a particular species leads to an image with 50 different species on it; on the other hand, the alternative would seem to involve cropping the original image so that only the relevant species appears for use in EOL. Is there some sort of protocol for dealing with this situation?

In any case, thanks for this wonderful resource!

BHL said...

Thanks Jennifer, great question. Indeed, it is not the most useful for EOL species pages to have plates linked in from BHL that have multiple species on them. What is recommended is that you tag the page at a higher taxonomic classification in order to capture everything on the page in general. Instead of tagging species by species, try tagging at the family level, "taxonomy:family=Acanthopteroctetidae" or even the order level, "taxonomy:order=Lepidoptera" if need be. We are not in the practice of cropping our images as this would disturb the integrity of the page image and it would take a lot of work to crop and re-upload. Should you want to experiment with cropping our images, please feel free! Thank you for reading our blog and thanks so much for taking the time to tag our Flickr images. We really appreciate it.

Anonymous said...

I would disagree with the direction of cropping the images & providing alternates. Scientists using EOL are quite used to the practice of including several organisms on a single plate. An alternative would be to draw a note around each organism & leave a machine tag as a note

BHL said...

That's a great suggestion Anonymous. To date, no one has created any notes for our images in Flickr. Sounds like a better alternative to cropping. Please feel free to add notes with tags when you can. Thank you for taking the time to leave your comments.

Jennifer W. Hanson said...

Anonymous, thanks for the point about notes. BHL, you are about to get a lot of notes on your set of plates from The Moth Guide! :)

Daniel Mietchen said...

I find it irritating that BHL images on Flickr are labeled CC BY-NC-SA. If they are actually PD and that option is not available, wouldn't it make sense to either add a PD statement with each upload, or to use the option with the least restrictions?

Besides, have you looked into adding materials to http://commons.wikimedia.org/wiki/Category:Files_from_the_Biodiversity_Heritage_Library ?

Martin said...

Daniel, thanks for the comment. Just yesterday we were working on a comment that we can add (either to the profile and sets or the actual image) that clearly call out the images as PD.

We were attempting to choose the least restrictive license offered by Flickr (which is not a full CC license option). Unlike content in the Flickr Commons, we're unable to use "No Known Copyright".

And YES! We love the Wikimedia BHL content. We're working with Wikimedians to try and streamline the process. One thing that Flickr allows is crowdsourcing identification of images so that they're easier to move over to Wikimedia Commons. We're have more information on this by summer.

BHL said...

Some guidelines from our EOL colleagues:

"Flickr notes display only on Flickr itself. That text does not get imported to EOL. If you have plates with multiple species, you should add binomial machine tags for each of the species featured. You can also add left to right, etc. information to the image description. One thing that's important: Do not mix taxonomic levels in the machine tags when there is more than one organism represented on a plate, i.e., if you know the species for one organism and only the genus for the other, do not add a binomial tag AND a genus tag. Our harvester cannot handle this. Currently, the only way to deal with images like this is to upload them twice to Flickr and to tag each version of the image individually.

At some point, I hope people will take the initiative to crop individual images out of these plates and re-tag and re-upload them to EOL. This kind of thing is happening a lot on Wikimedia Commons."

Many thanks to all those contributing to this dialogue and for tagging our Flickr content!

BHL said...

Some guidelines from our EOL colleagues:

"Flickr notes display only on Flickr itself. That text does not get imported to EOL. If you have plates with multiple species, you should add binomial machine tags for each of the species featured. You can also add left to right, etc. information to the image description. One thing that's important: Do not mix taxonomic levels in the machine tags when there is more than one organism represented on a plate, i.e., if you know the species for one organism and only the genus for the other, do not add a binomial tag AND a genus tag. Our harvester cannot handle this. Currently, the only way to deal with images like this is to upload them twice to Flickr and to tag each version of the image individually.

At some point, I hope people will take the initiative to crop individual images out of these plates and re-tag and re-upload them to EOL. This kind of thing is happening a lot on Wikimedia Commons."

Many thanks to all those contributing to this dialogue and for tagging our Flickr content!

Chris Freeland said...

But here's an example of where cropping won't work:
http://www.flickr.com/photos/biodivlibrary/6505684989/in/set-72157628403496467

I would prefer that the images are tagged as accurately as possible and that hopefully we can find a way around the limitations of EOL's ingest process. These images are of use to many communities.