100 PURCELL AND AOL - UPDATE
September 17, 1997
In July we reported that the "Pictures of the World" site of Carl and Ann Purcell had been dumped by AOL.
It turns out that while their contract was terminated June 30th, and their site on AOL's Travel Channel closed, it was possible for AOL users to get free access (without any compensation to, or normal control by, the Purcell's) to their entire 8,000 image library for 22 days in July. This was accomplished by using hot links from other AOL sites.
The fact that these hot links were left intact, when AOL could have easily cut them, resulted in over 18,500 copyright infringements during that period.
The Purcell's have filed a multimillion-dollar copyright infringement lawsuit in federal court in San Francisco against AMERICA ONLINE. This case may turn out to involve more infringements than any other case in history.
One of the alleged violations is under the keyword, "contest." A small icon shows Steve Case and Ted Leonsis titled "the legal guys" and when the user clicks on that icon Purcell's copyrighted picture of the Supreme Court is displayed. At this site the user is warned about the importance of observing copyright. But, AOL failed to observe Purcell's.
September 17, 1997
Understanding how words and phrases are used to do digital searchs for images is extremely important for photographers as they determine which agencies and which systems to align themselves with in the future.
Basically, there are two systems, one with the capability of narrowing the search when you add words (keyword), and the other which tends to broaden the search and provide additional hits whenever defining words are added.
The systems used by PNI, Corbis and Index Stock's "Photos To Go" database are of the broadening type. PNI calls their's "Natural Language Search" and Index uses the term "Free Text". Index's "TeleFocus" site uses a keyword system that results in a tighter selection when additional defining words are added.
The first thing you need to recognize is that with most systems the words that identify the pictures are written in a block of text one word after the other. The search engine simply looks at the entire text block and determines if any words match the search request.
Assume you are doing a "Natural Language" search for "boy with a dog." Images with the highest number of commonalities are shown first. If an image happens to have all four words in its word list it will be shown first. It will be followed by images that have three words, two words, and then one word of commonality.
Consequently you may get images that show boys without dogs, or dogs without boys. Also, because these are very common words you could get a lot of pictures to look through, none of which would really fit your search requirements.
If the first pictures that come up don't show a boy and a dog there probably won't be any "boy with a dog" pictures in the group. But, if it says you have 500 matches for your request there is always the feeling that the picture you want maybe somewhere buried in that group. There is the tendency to want to check out all the hits. This can become time consuming and discouraging.
Let's suppose that you did find one picture of a teenager (boy) with a dog, but what you really wanted was a much younger boy. If you search for "little boy with a dog" you have broadened, not narrowed, your search. Now, in addition to the images you got before you may get "little fish" or a "little hat" or anything where the adjective little was used to describe it.
One of the features of the PNI search engine Kodak acquired is that in addition to finding the exact word you request it tries to figure out what you really want. For example, if you asked for a student it might also search for the word school, or boy, or girl. Students are usually in schools and are probably boys or girls. This can result in many more hits, but not necessarily images you want. The user has no idea what other words the software is looking for except to try to discern this from the images that appear.
This can be very frustrating.
Keyword search looks for the exact words that are input. If you use a phrase like "Yellowstone Park" it will only find images where that exact phrase was used. It will not find all parks nor "Yellowstone National Park." If you input "boy" and "dog" it will only show you images that have both "boy" and "dog" in the word list for that image.
With this system it is much easier to narrow searches. It is my belief that as the number of images in any particular database increases the ability to narrow searches will become increasingly important for the user.