Keywords: Hierarchical or relational system?

Status
Not open for further replies.

PatrickC

Active Member
Joined
Jan 30, 2008
Messages
192
Location
UK
I began keywording our images when there were no hierarchical keywording systems available, and you were expected to type in keywords from the top of your head.

I developed my own system, which I refer to as relational, based on a Word document with lots of internal hyperlinks. I wrote some macros to increase my productivity.

The main document consists of lines of keywords which are related, organised in a way which makes sense for our library. I very much doubt if it would work for anyone else, but it grew organically and is good for us.

The first page is only hyperlinks which lead into sections, like geographical locations, people, environment, education, culture, economic activity etc.

When I get to some keywords I want to use, I click in the left margin to select the line, or double click individual words. I've set up a right-click menu item 'Copy Keyword', which copies the selected words to a second document.

I navigate the main document using an enhanced version of the web toolbar to go to the beginning of the document or back to the previous location, and I also use embedded hyperlinks to move to related areas of the document. For example, on the 'People' page there are links to 'Emotions', 'Lifestyle', 'Sexuality' and 'Indigenous' (amongst others).

The toolbar also has an icon 'Finish Image', which saves the main document (thereby saving any changes I have made, enabling it to grow organically and retain any flashes of inspiration I have while working on a specific image), prepares the keywords in the secondary document into a format suitable for Lightroom to use, copies the keywords to the clipboard, then clears the second document ready to keyword the next document.

I then just paste the prepared keywords into the Keyword Tags box in the Keywording pane (RHS) in Lightroom.

While my system was growing, commercial hierarchical systems were increasing in popularity and becoming more widely available. But I stuck to my system because a) it was adapted to our specific needs, and b) I have a strong suspicion that hierarchical systems lead to inappropriate keywording.

Why? A hierarchical system assigns 'implied' keywords, which are 'parents' of the actual word you select. So, for example, selecting 'cat' might add 'feline', 'felis', 'felis sylvestrus catus', 'mammal', 'animal'. But imagine the confusion if the 'cat' referred to is a catfish? There is even, believe it or not, a fish known as the bumble bee catfish, which is in a group known as the wasp catfish!

Some systems allow you to deselect parents, but if you are in the process of keywording thousands of images, it is all too easy to forget that you are assigning these implied keywords, because you don't see them on the screen as you work.

In my system I always have to select the words, and even so if I'm tired or under pressure I occasionally select a whole line which includes some strays.

My system results in a flat file of keywords, which is fine if the software just keeps the actual words in a database which relates to the image file, or, as in the case of embedded or sidecar xml, in a file in a 1-to-1 relationship with the image file. But it breaks down, as I have discovered to my cost with Lightroom, if the database relationship is different.

I don't understand exactly how keywords are linked to images in Lightroom, but I'm aware it isn't in the kind of relationship in the paragraph above. Instead of assigning a keyword to an image, I believe you are doing the reverse; assigning an image to a keyword. This, if I'm not wrong, enables searching for images to be faster.

It seems that I'm being driven into the hierarchical system, just to make my catalog useable. But, because I've never used one, I don't feel confident that I understand the structure well enough to design a hierarchy I won't later regret.

How many levels are optimum? Is it OK to have the same word in several places (eg cat in mammals, cat in fish)? If yes, how does Lightroom distinguish between instances of the same word? If yes, what happens when I import already flat-keyworded images into a catalog with a hierarchical system? How do I prevent Lightroom from assigning the wrong instance? How best to avoid assigning inappropriate parents? (My kids might be interested in that one :D).

I'm sorry for the length of this post, but I think this is a discussion worth having.

Patrick Cunningham
 
I don't understand exactly how keywords are linked to images in Lightroom, but I'm aware it isn't in the kind of relationship in the paragraph above. Instead of assigning a keyword to an image, I believe you are doing the reverse; assigning an image to a keyword. This, if I'm not wrong, enables searching for images to be faster.
There are different ways to link images to keywords. Selecting one OR MORE images in the library and dragging them over a keyword (Left pane). Dragging a keyword from the left pane to selected image(s). Select image(s) and typing comma separated keywords in the right pane.
How many levels are optimum? Is it OK to have the same word in several places (eg cat in mammals, cat in fish)? If yes, how does Lightroom distinguish between instances of the same word?
That should not be a problem as you probably have your cat fish images in a different folder than your cat images. Filtering on "cat", you can also select just the relevant folder(s).
This unless you are catch catching a cat fishing a catfish! ;)

I myself use only three levels with the parents tagged for not being exported. So I leave comments on levels for more experienced users or those using many.
 
Patrick,

I've been using a hierarchical keyword scheme for a couple years now and really like it. I think there are a couple misconceptions that you have about the way LR keywords are stored. I believe the keywords are assigned to images in the database, not images to keywords. It really doesn't make any difference externally. If you assign keywords to the images and then generate an index on keywords, you have the equivalent (from the database perspective) of assigning images to keywords. If you save your metadata to your image (Ctrl-S) and then look at the image metadata with an external tool, you'll see that the keywords are all there, including the hierarchy.

I also don't understand your comments about "inappropriate parents". If you construct your hierarchy correctly, I can't understand how you would ever have a hierarchical parent that is inappropriate to one of its children. Why would one ever assign the keyword 'cat' to a catfish? It's a fish, not a cat. The hierarchy needs to be built on relationships between parent and child, not on etymology. (unless that's your need, in which case maybe it would make sense to have the word catfish under cat).

I find it very useful to be able to find all of my animal pictures just by clicking on the 'Animal' keyword in the hierarchy. On the other hand, if I'm looking for a specific picture I remember taking of an alligator, I'll click on the 'Alligator' keyword which is a couple levels down in the hierarchy, and I'll only have to wade through all of my alligator pictures.

You do have to spend some time thinking about the structure of your hierarchy before you start. I think it's very important that your top level be pretty much orthogonal. By that, I mean top level keywords that are completely independent. For example, my top level includes keywords like, Location, Status, Subject, etc. Further down in the Subject hierarchy, I have an Animal keyword; somewhere below it, I have Cat and Dog. I also try to limit the amount of overlap between portions of the hierarchy. If I had a Pet keyword, I would not put a Cat keyword below it. If I did, I would have two places to look for cats, which would only be confusing. Instead, if I need to look for images of cats that are also pets, I use the combination of the two keywords.

Hope this is of some use. I think once you start to use the hierarchical system, you'll really like it. As an aside, I never use the RH keywording panel in LR. I select groups of images in grid that need the same keyword and drag the entire group to the appropriate child keyword on the LH keyword list. I find this a very effective way to work for keywording large groups of images.
 
I am looking at keywords from the POV of a picture researcher who doesn't know the collection. Many people would put "cat fish" into a search engine instead of "catfish", so it's important to have the word "cat" on the image of a catfish. Because I'm preparing images to be searched in this way, the fact that the cats and the fish might be in different folders is irrelevant - in fact I file my images only by date, so they are likely to be ing the same folder.

But this is only one example. There are plenty of occasions when a single word can be relevant in several completely different contexts, when one set of parents are incorrect and another are appropriate.

I know that I can write a flat file of keywords to an image - which is what I do when sending them to an agent at the moment. What I was trying to find out is the underlying mechanism of Lightroom - but I take Bruce's point that it is to a large extent irrelevant.

It may well be that using the Lightroom system will be quicker. One of the failings of my system is that it doesn't readily lend itself to keywording batches of images. While I was working on scans of our back catalogue, that didn't matter much because it was rare to have a sequence of similar images, but with digital originated images the situation is different, and the system built into Lightroom may well work for me.

Thank you too for the explanation about your orthogonal top-level, that is a very helpful idea. I'll get thinking again.....

Patrick Cunningham
 
I also don't understand your comments about "inappropriate parents". If you construct your hierarchy correctly, I can't understand how you would ever have a hierarchical parent that is inappropriate to one of its children.
But I think this may sometimes happen. For example, I have a parent keyword named "Names". Under it are "Firstname Lastname" of individuals. In this case, I do not want the word "Names" to appear in the exported files so, I set them accordingly with the "Do not include at export" option. Still think it was a good idea to use "Names" so I can collapse when not needed.
 
But I think this may sometimes happen. For example, I have a parent keyword named "Names". Under it are "Firstname Lastname" of individuals. In this case, I do not want the word "Names" to appear in the exported files so, I set them accordingly with the "Do not include at export" option. Still think it was a good idea to use "Names" so I can collapse when not needed.

Yes, I do the same thing with many of my top level keywords. I don't see them as inappropriate, so much as I see them as unnecessary to a third party. In other words, they're not wrong, they're just not useful for purposes other than to make the hierarchy make sense. Now, I'm not sure that this makes sense.:?
 
Who's using the Controlled Vocabulary hierarchy, and what are your views? It's not cheap, and I can't decided if it is appropriate for me.

I've just had a quick look at the CV site and I can't find anything about the differences between US 'English' and English English. If I'm going to be forced to use the bastardised US version with English only offered as synonyms I'm not interested. Not that I've got anything against US English, it's just that I'm English....

Patrick Cunningham
 
Last edited by a moderator:
Who's using the Controlled Vocabulary hierarchy, and what are your views? It's not cheap, and I can't decided if it is appropriate for me.

I've just had a quick look at the CV site and I can't find anything about the differences between US 'English' and English English. If I'm going to be forced to use the bastardised US version with English only offered as synonyms I'm not interested. Not that I've got anything against US English, it's just that I'm English....

Patrick Cunningham

I looked at the CV and decided that I would spend enough time modifying it to reflect my needs that it would be just as easy to build one from scratch. Mine is pretty organic and still growing. I started with some top-level keywords that I knew I would need and then just let the hierarchy develop as I saw a need for it. Just added a few more terms last night when I realized one of my categories had too many images that couldn't be separated by keywords. It's simple to do and has the advantage that I didn't have to invent the whole thing before I started using it and I only have keywords that I find useful. However, I only keyword for my own uses, not for the public, so YMMV. Cheers,
 
Status
Not open for further replies.
Back
Top