Sonny Software

I begin to appreciate the elegance and intuitiveness of tag clouds, a feature that I had previously underestimated.

Here are a couple of suggestions:

1. For each tag in the tag cloud, ignore any characters following an apostrophe (and the apostrophe itself).
This would ensure that Nobel and Nobel's are counted as one and the same tag.
But it would not solve the problem of singular and plural forms, which is more complicated.

2. [More general solution] Allow the user to combine several tags into one.
For example: Symphony, Symphonies, Symphony's, Symphonies' could be aggregated into Symphony.
One could also aggregate such variants as Marx, Marxist and Marxian.
In terms of implementation, one would perhaps drag-and-drop the 'superfluous' tags (Symphonies, Symphony's, Symphonies' ) onto the combined tag (Symphony).

Interesting idea. But it would require intensive user intervention because Bookends anything about English or its conventions (not to mention non-English languages). So you'd have to curate your own database. Not a problem for a few dozen terms. But it would become burdensome if you really wanted to make use of a feature like that...

Jon
Sonny Software

Yes, it could conceivably be a bit burdensome for clouds with more than 500 tags, but let's emphasize that this would be a 'transparent' and optional feature that would not interfere with clouds as we now know them.

Also, tags to be merged would invariably be consecutive (e.g. Religion, Religions, Religious, Religiousness), so the refinement process would only involve a single pass at most. Easy to do because consecutive words sharing the same stem stand out visually.

Currently, because variants of the same tag cannot be merged, the relative sizes of individual tags can be visually misleading. To give a crude example, in a database mostly about religion a "Medicine" tag may be larger than the "Religion" and "Religious" tags individually, even though the total number of occurrences for the latter two tags may be greater.

How would you display merged tags? String the words together (which could be pretty ugly if there are more than a few), e.g.: religion/religions/religious/religiousness/Christianity/Christian/Judaism/Jewish/...

Or create pseudotags in preferences, where you would define a tag (e.g. Religions) and manually add terms to that. Then have that tag displayed if any of the terms are found?

I'm not saying that merging terms might not be useful, but I'm seeing issues with implementation.

Jon
Sonny Software

Hmmm… how about giving each tag a contextual menu (right-click), which would include a list of all aggregated sub-tags. Clicking on one of them would simply remove that sub-tag from the aggregate and move it back into the tag cloud as a standalone tag.

I think that simply a small asterisk or plus sign on the upper-right corner of such "aggregate" tags could be used to visually distinguish them from single-term tags. I agree with you that displaying a chain of tags in the tag cloud ("computer-computers-computerized") would look awful.

Just an idea…

Nice ideas. You could of course also try an elipsis: … but then again it might take too much space.

To take this further, because it is managed by the user, tag aggregates or clusters could not be limited to similar forms of the same word (e.g. singular vs. plural), but would allow for logical (or illogical) groupings of terms (e.g. "red", "blue", "green" under a tag "colors…"). So I would lean toward keeping the display of individual tags in addition to showing the tag cluster. I'm thinking that you would enter tag cluster names in preferences, along with the words to be aggregated. You'd also be able to add to or remove terms from a cluster by right-clicking in the tag cloud view. In the example above, you'd see something like this in the tag cloud view

blue 6
colors… 10
green 1
red 3

And if you click on Colors of course the list would be filtered for the 3 terms it contains.

Jon
Sonny Software

I agree that enabling arbitrarily named large-scale tags for grouping individual tags (e.g. "red", "green", "blue") into broader concepts ("colors") is a great idea.

My concern with your suggestion, however, is that it does not address the visual and conceptual confusion of having separate tags for different forms of the same word ("King", "Kings", "King's").

It boils down to a matter of taste, I guess, but personally I would prefer a neat and visually consistent tag cloud, in which tag sizes accurately reflect the semantic prominence of tags in the database, to a tag cloud with arbitrarily nameable clusters.

That said, your suggestion could still work excellently provided that:

1) it is possible to use the name of a single tag as the name of the tag cluster containing it, for example:
King 6
King's 2
King 3
Kings 1

and

2) it is possible to "hide" a tag from the tag cloud (e.g. by means of a contextual menu option) without actually withdrawing it from the underlying calculations.
King 6
King's 2 [Hidden]
King 3 [Hidden]
Kings 1 [Hidden]

Now couple this with a way to create such "tag clusters" using drag-and-drop (i.e. without recourse to the "preference pain"), and I think we have a killer feature!

Right now #1 would be possible (except King will be King…) and #2 would not. I'm thinking of using this just for word tags (e.g. from titles, abstracts, and notes), not term tags like authors and keywords (the difference is that the format can't have spaces, of course). As for the drag and drop, no, but I think tags could be added/removed with right-clicks, which should be easy (just like you can add terms to the stop list right now). Oh, and I'm thinking of using the term "virtual tags" to describe tag aggregates.

Jon
Sonny Software

Just to clarify—with apologies for this barrage of posts: What happens if you remove a single tag from the cloud, but that tag is a member of a virtual tag? If removing a single tag from the cloud does not exclude it from the summations for the virtual tag, as I hope, then I don't think we are in disagreement here.

At any rate, thanks a lot for considering this feature.

You mean if you add "you" to the stop list of words but specify it will be part of a virtual tag called "pronouns"? Hm, I think the stop list wins. The word will not appear in the virtual tag (and if the only word assigned to the virtual tag is "you", the virtual tag will not be created).

I'm still not decided on whether to allow terms in the virtual tags or not (e.g. authors). I'm going to have to play with it and see how it feels. I'm also not sure whether hiding terms in a virtual tag (as you want) or not is a good idea (I can see it's reasonable for the singular vs. plural forms you mentioned, but don't think you'd want "green" hidden if you had a virtual tag for "Colors"). Having options to treat different virtual tags and words/terms differently just gets messy, so I want the same behavior for all. Since I think the real benefit from a feature like this is logical grouping of tags (not consolidating singular and plural forms, which requires a lot of work by the user, since there are thousands and thousands of words that would fall into this category) I'm leaning toward leaving them all in (the tags and the virtual tag).

Jon
Sonny Software

I agree with the original poster that tag clouds are great. I basically moved all my primary source notes to BE and tags help me navigate through them very well.

I'm intrigued how you'd implement this--as hierarchies/relationships between keywords? or at the tag word level? Anyway, it's potentially very useful, especially if it's easy to use. Some suggestions below:
- maybe each tag can show a mini-button, +/-, next to it when the mouse hovers over the word (like the x in Safari's tabs or in email addresses in Mail.app). where + expands the cluster, - removes it, or something like that. Or maybe just a little gear, that the user can click on and get options. It would be really slick to drag and drop to automatically create a cluster, but I see how this can be very difficult to implement and may result in something confusing.
- I'd rather have a "Cluster" vs "Atomic" pill kind of button, instead of having to chose word by word if they are shown as clusters or as individual tags.

And since we're talking about tag clouds, it'd be useful from my perspective to be able to choose arbitrary sets of fields to show in the tag cloud. E.g.: abstract & authors. Perhaps this can be set in the the Tags tab in Preferences.

Thanks!

The implementation I'm considering is simple. You can create virtual tags with any name you want, then add words to that tag. When displayed in a tag cloud, the virtual tag name will appear with an ellipsis (e.g. Computers…), and if you click on it the reference list will be filtered by the words the virtual tag "contains". You'll be able to add/remove tags from the virtual tag in preferences or in the tag cloud view with a right click (contextual menu).

Undecided at this moment is whether authors/editors/keywords (e.g. terms) can be added to a virtual tag, or whether that will be restricted to individual words (i.e. the tags generated from the abstract, notes, and/or title).

Also undecided is if a term moved to a virtual tag will then be removed from the tag cloud display or left there (so you'd have "computer" included in the Computers… virtual tag AND also displayed as it's own clickable tag).

Finally, you can't mix and match fields. Authors/editors/keywords tags are terms, meaning they can contain punctuation, spaces, and always take up one line. Abstracts/notes/title tags are individual words that cannot contain punctuation or spaces, and rarely would take an entire line. Bookends has to know which kind of tag you click on to generate the appropriate search.

Jon
Sonny Software

Jon wrote:Also undecided is if a term moved to a virtual tag will then be removed from the tag cloud display or left there (so you'd have "computer" included in the Computers… virtual tag AND also displayed as it's own clickable tag).

Perhaps it would be best—if possible from the developer's perspective—to leave this decision up to the user (via an option in Preferences)…?

As I noted earlier in this thread, it might get very messy if I do that on a tag-by-tag basis. Perhaps it could be an all-or-none choice. All of these issues may have to be rethought when I actually get around to trying to implement something.

Jon
Sonny Software