Keywording : Best Practices to Accommodate Distributors

CEPIC Seminar

By David Riecks.

This CEPIC session included key personnel from all of the major stock image distributors, Getty Images, Corbis, Jupiter Images and Alamy. The focus of the presentations and discussion was on how contributors can best prepare their images for submission, especially in regards to keywords. Keywords are essential for delivering a good search experience for customers and in some cases they are the only word terms that are searchable. The Alamy, and Corbis sites do allow visitors to search the caption, but don’t weigh the results as heavily as the keywords…the Jupiter site not only allows caption search but gives this field more weight.

Mary Forester: Director of Search Data Strategy for Getty Images explained that they have a 46 page document called the “Getty Images Guidelines for Keyword Providers of Creative Content” that explains all the details. Coverage of basic areas, location, concepts, people and ages of models, and objects is expected; however they should not  include tangential, speculative or opposite keywords.  Synonyms are backups to the main terms, and also include translations into other languages. They expect 100 percent compliance; and content will be rejected if it doesn’t meet requirements.

Andrew Labonte: Senior Manager of Search Metadata at Corbis said they also have Keywording Guidelines for contributors. They currently do most keywording in-house and are using a hybrid, “Mapalogging” system. This allows them to get about 1000 images cataloged a week per person, rather than 400. They do review partner images and remove keywords for those that are not appropriate, which at present is about 39 percent of these Media Partner keywords.

The types of keywords needed vary by the type of photographic stock. If editorial content then captions are essential and the date is important. If travel imagery, then the location information is of paramount importance and date is not as critical.

Alan Capel, Alamy, Head of content for Alamy said even he has to check the site to see how many images they have. At present that amount is about 12 million pictures. Unlike the other distributors, Alamy requires that contributors do all their own keywording and captions and they do not use a Controlled Vocabulary. There is substantial competition as everyone is doing their best to have their images seen.  The actual sort order of images on site is driven by customer activity.  If your image appears in the search results and is ignored, you lose. If it appears and is clicked on, you win, and if it is licensed it you really win.  

They have introduced three different keywords fields, essential, main and comprehensive in order to help weight which words are most important. Essential keywords are weighted very highly, followed by the main keywords, the comprehensive and the caption field. The click-through-rates are measured for all pictures and are also used to help determine which keywords are most relevant.  In order to keep things running smoothly they won’t ask you to update an image, if you exceed the specific character lengths, instead your image will be rejected.

Janice Lodato: Director of Search Data for Jupiter Images noted that they currently prepare images for one of eight or nine search engines. She described how they keyword, starting with the “Aboutness” of an image, and then expanding to include concepts, age and ethnicity, appearance and action, and location information. They ask that their partners use a simple relevance test and ask themselves, if they were searching for “X” term would they want this image in the results.  In their future she sees a need for normalization, standardization and more automation.

Kirsti OSullivan, owner of Keywording.com, gave some great tips on how to use special spreadsheet functions to ease the pain of preparing submissions. Functions such as concantenate, allow you to join together columns of data, whereas the LEN option allows you to know the exact character length of a field. There were audience questions regarding differences in how distributors deal with American and UK spelling, conceptual keywords, building controlled vocabularies, and the use of IPTC metadata fields for input into systems.