Rescuing Bad Keywording - It's Normally Mission Impossible

We are often approached by people wanting bad keywording made good, but almost without exception they end up having the keywords redone from scratch.

Apart from minor improvements or enhancements, "fixing" bad keywords is normally more trouble than it's worth. There are several reasons why:

1. Problems are rarely uniform. If there were just a few things to be fixed, and the problem was the same for every image, it would be a relatively efficient process to look for what needs fixing. However, if there are numerous problems in no set pattern then each image has to be reviewed in detail and a strategy worked out for each one, which is very time-consuming.

2. Keywords have to be removed as well as added. When keywording from scratch, pretty much you are only adding keywords. When fixing images you normally have to remove bad keywords (after idenitifying them first), which means a second process.

3. You can't trust the spelling. When images are badly keyworded it is difficult to know what can be relied upon. Whilst our own databases are constantly checked and rechecked to make sure place names, proper nouns and other words are correctly spelled, if a bad keywording job has been done you inevitably find yourself being drawn into doing a lot of checking of spelling because you're not sure if the same rigour was applied the first time the keywords were put in. Getting it right first time is very much faster than having to go to a dictionary or do research on the internet.



4. The keywords aren't the right standard. Sometimes the problem is that keywords need to be changed to meet particular keywording standards, such as Getty. This means doing things like finding and taking out plurals or assessing if particular words are acceptable for the standard. There are also problems with meeting requirements for the number of keywords, which can lead to lengthy delays working out which keywords need to be jettisoned.

5. Spam, spam, spam, spam. Sometimes images have been horrendously over-keyworded. If an image contains 300 words, and needs to be trimmed to about 50 relevant terms, it's obvious that mathematically you're better off putting the correct 50 in than working out which 250 you're best to get rid of.

6. When they're bad, they're very, very bad. Often the sheer volume of problems multiplies the time needed to fix the keywords, for all the reasons above.