Brussels / 31 January & 1 February 2015

schedule

Interview with Jonas Öberg
Automating Attribution. Giving credit where credit is due

Jonas Öberg will give a talk about Automating Attribution. Giving credit where credit is due at FOSDEM 2015.
Q: Could you briefly introduce yourself?

On my 22nd birthday in 1999, I received a phone call from Richard Stallman to tell me the root password of the GNU project web server. I’d been a free software volunteer for some time before this, and continue to be to this day. In 2001, I co-founded the Free Software Foundation Europe and served as its vice president for many years. I’ve organised conferences (FSCONS) and worked as regional coordinator for Creative Commons in Europe and Central Asia. I’m a Shuttleworth Foundation Fellow and as part of my work for the Foundation, I launched Commons Machinery to focus on metadata for creative works.

When I want to avoid technology, I like to say that I’m renovating a house in northern Sweden. That’s only partly true today; we’ve always had Internet (ADSL and mobile) at that house, but we leveled up this past summer with an installed fiber connection. We can now get gigabit network at the summer house, and just 28 Mbit ADSL where I otherwise live outside of Stockholm. So much for avoiding technology during vacation!

Q: What will your talk be about, exactly? Why this topic?

When I worked with Creative Commons, I saw at first hand the practical problems people have attributing works accurately. Sure, one or two images may be fine, but when you’re juggling even tens of images in a presentation, it’s easy to lose track of what came from where. Thinking about this more, I realised that not attributing accurately is even more troublesome than just violating a CC license: it makes images lose the context in which they originally appeared, and when we can’t see where an image comes from, or who authored it, it loses some of its meaning, and it becomes less valuable to us.

I’ve come to consider attribution to be even more important than copyright, and I believe that if we can develop the tools that we need to attribute easily and fairly, we don’t actually need copyright, and we wouldn’t need licenses. My talk is, by extension, of a post open licensing society, but building towards that, I will talk about the technology we’ve developed from Commons Machinery to figure out what we can and cannot do automatically, and introduce a range of libraries and tools (all on GitHub) that has come out of this work, including (which I’m most excited about) our algorithms for comparing images and searching a database for similar (not identical) hash values.

Q: What do you hope to accomplish by giving this talk? What do you expect?

Just two months prior to FOSDEM, we released the first useful tool to come out of Commons Machinery: Elog.io, our catalog of digital images and corresponding browser plugins to look up information in that catalog. With this, we’ve come to a natural conclusion of the first part of our journey and shown that what we imagined works in practice. We’re preparing ourselves to take the next steps, which involve scaling our technology to include more images, but also turning the catalog into a read/write storage and making it into a community curated metadata storage.

This will be an exciting challenge, so I’m hoping that my talk at FOSDEM will bring some valuable feedback to us in terms of our current thoughts and directions. Is what we’re doing useful? Are we going about it in the right way? Do the challenges have other solutions which we haven’t thought of?

This talk will also be the first time we present the standalone libraries that have come out of our project so far, so I hope that raising awareness of them during my talk will also encourage additional use of them.

Q: What’s the history of the Commons Machinery project? Why was it started and how did it evolve? Has it become what you planned it to be?

When I started Commons Machinery in 2013, I did so with the explicit aim of using embedded metadata as a way to solve all the world’s problems. Since then, I’ve come to realise quite a lot about metadata, including that embedding it never really works as you would expect. From the Embedded Metadata Manifesto (a project by the International Press and Telecommunications Council) we’ve learned that most social media sites today strip metadata when images gets passed through their networks. In our own work, we also discovered that the same is true for a lot of the applications we use daily. Support for metadata is really missing.

So Commons Machinery became an experimental factory where we, thanks to the support from the Shuttleworth Foundation, could spend a year and a half feeling our way forward and develop a number of throw-away prototypes in the process. We’re now coming to an end of their support, so we’ve had to wrap up a lot of the activities we were doing, and while it would be fun to keep experimenting, we believe we have a good understanding now of what needs to be done. Now it’s just about doing it.

Q: The Elog.io browser plugin is an interesting tool to view licensing information of images on websites you visit. Do you plan to add other sources than Wikimedia Commons to the Elog.io catalog?

We will definitely add more sources. During 2015, we’re looking at including images from Europeana, Safe Creative, and Flickr. We have discussions with all of them, so I’m convinced we’ll be able to add parts from their collections. This will result in an Elog.io which is able to match more and more images that you find online, and more and more accurately.

One of the challenges there is to add these sources quickly enough! When we did our initial harvesting of information from Wikimedia Commons, we did so at a rate of ca 2 million images per day (23 per second!), and even if we manage to keep up the same rate when harvesting Flickr, it’d take us more than five months to work through the 305 million images with a CC license.

Q: Which other new features can we expect this year in the Commons Machinery’s software?

A stretch goal for 2015 is if we can flip the read/write switch already now, but that would likely not be until later this year if it happens in 2015 at all.

Q: How can interested people help?

We need code, ideas and networking. For some of the libraries we’ve developed, there are some ideas on how to improve them further. For instance, over the holidays, we got a proposal for an adaptation of our image hashing code that would theoretically generate more unique hashes with higher resolution. This would be excellent, but we haven’t gotten around to testing it yet, and we would need someone to branch the code and implement the alternative algorithm and setup some tests against our sample images. The same is true for a lot of other libraries and utilities we have too.

And, just as with the idea on how to improve our hashing code, we always need people who can look at what we’ve done and give ideas for how they can be improved further. It doesn’t need to be something you can implement yourself; just providing your ideas for improvements is a valuable contribution in itself. When FOSDEM rolls around the corner, we’ll also be just days away from launching a crowdfunding campaign to ensure that we have some funding to hire people in the coming year too. Help is always needed in spreading awareness of that campaign, and our of work in general.

A special shout out to those who are working in organisations where they have collections of images openly licensed or in the public domain: we’d love to be able to include those images in the Elog.io database too, so we’d appreciate hearing from you.

Q: Have you enjoyed previous FOSDEM editions?

The thing I enjoy about FOSDEM is that even if I hadn’t enjoyed it (I did), I would still go. Because it’s FOSDEM. You go to FOSDEM. It’s as simple as that. But truly, FOSDEM is always a great experience. I enjoy going there every year, and I hope to have the chance of continuing to go there every year, even if in a few years, it’ll need to be with my son accompanying!

Creative Commons License
Creative Commons License

This interview is licensed under a Creative Commons Attribution 2.0 Belgium License.