DISQUS

Mat · 1 year ago

The idea is that the correct tags will be generated more frequently by the crowd, thus giving those terms a greater preponderance than the mistakes or falsehoods... simple enough, no?

mathewi · 1 year ago

Definitely very simple -- and theoretically it should work too.

Leila Boujnane · 1 year ago

Matthew: this is really fantastic! from Matt Raymond, the Library of Congress chief blogger: "One of our goals, frankly, is to learn as much as we can about that power simply through the process of making constructive use of it." I believe that people with interest in these photographs will tag them best they can, the Library of Congress will review the tags as their stated goal is to see if these tags and classification can actually feed back into the Library of Congress catalog. We trust they won't be doing this blindly!

abacab · 1 year ago

Google's way of pairing people up using their Image Labeler service to agree on tags for images seems a reasonable method as well... but I agree that the bad tags will basically fall away purely statistically.

Will Pate · 1 year ago

I'd love to know what the statistical rate of incorrect tags will be. It would be easy to have a mechanism for voting a tag as "incorrect". I bet it would be very low, but that's just an educated guess.

mathewi · 1 year ago

Yeah, maybe I'm being overly negative on this one.

geoperdis · 1 year ago

Mathew:

As you probably already know, organizations like the Library of Congress, wire services and stock photo agencies currently use controlled vocabularies ( highly structured and deep taxonomies) to classify the photos. This is a tried-and-true way of cataloguing photos (and other media) for easy retrieval, assuming that the user searching for photos has a basic understanding of how the photos are organized and structured in the first place.

But if the person searching for a photo is not a pro (photo editor, art director, researcher, librarian, etc.) or already familiar with the vocabulary structure used to catalog a set of pictures, they'll probably opt to use a free text or natural language querry (aka google it) to try and find the photos.

By putting the collection of photos on Flickr, the Library of Congress has a unique opportunity to poll a wide range of people and get feedback on how they see or identify the photos through tags and comments using a free-form, non-controlled cataloguing method. And the Library can use those insights to perhaps improve their structured cataloguing and indexing process and terms to enhance search and findability in their picture collections.

So I think that this opportunity to see how non-pros identify the Library photos on Flickr far outweighs the potential for (likely small) errors in the tagging process. And having non-pros doing the tagging is exactly what you'd want here.

mathewi · 1 year ago

That's a good point. I hadn't thought of it that way.

robojiannis · 1 year ago

I actually find the crowdsourcing idea a great thing. It relies on the wisdom of the crowd. It's not about individual answers, it's about the collective intelligence. It is also a great deal of work; a work that Pro-Ams could do very well. A similar project of crowdsourcing and tagging was realized by the Metropolitan Museum.
If the collective tags are aggregated correctly, they could provide quality information, competing the data experts could provide.

Alexander Muse · 1 year ago

LOL - talk about tilting at windmills... 'someone might get a tag wrong!'

mathewi · 1 year ago

Thanks, Alexander.

Karoli · 1 year ago

Nah, I think you're stretching here. What do you mean by "correct tags" anyway? Shouldn't tags represent what the photo depicts or what value the viewer assigns to it?

Now, in your example with the typo, I really disagree with you, because typos a) happen all the time; and, b) often cause content to go undiscovered as the result of the searcher's typo and the content's correct spelling.

Tags with typos come in handy, more often than you'd think.

Ed Kohler · 1 year ago

The value of any photo's tags grows as more photos are tagged. For example, in aggregate, typos in tags can can understood and corrected for.

I suppose Flickr could create a certification level for tagging by professional taggers, but I don't think that's going to be necessary.

S fitz · 1 year ago

Search for the word "Brockton" in the Library of Congress photos and you'll see many folks from the city of Brockton chiming in and discussing the photos. They use Google Earth and MSN Live Maps Birdseye view to post current pictures of the locations discussed. That is something the LOC simply could never do efficiently.

mathewi · 1 year ago

That's a great example -- thanks.

Ryan Donahue · 1 year ago

Good points raised all around, and as someone who is neck-deep in a cultural heritage institution, I feel the commons allows us to harvest information that traditionally has been difficult to ascertain... perception of the masses, directly from the source.

While it's important to recognize incorrect tags, its also wonderful to be able to see exactly what is wrong. If enough people mis-label a car or a piece of equipment, we've just gained valuable knowledge (some people think this isn't a ford at all, but a chevy) .

It's very meta, in a way, but its interesting nevertheless.

Tracie Almond · 1 year ago

And these tags are not just for idle description. They are there to help people find material that would otherwise not show up due to the difference in professional (e.g. library, museum or archival) descriptive practices vs a layman's view. Therefore, things like typos in tags are probably no big deal. Sure they look a bit annoying if you come across one, but most people won't be searching by that, so probably won't even notice them.

Wise Startup Blog · 10 months ago

These small inconsistencies will definitely get worked out down the line, I think you're looking at just an isolated case of items and from a limited sample size.