Linking annotations together

[csillag]

I am creating this issue based on an earlier e-mail thread, because this is relevant to of our current discussions. (I am editing the text slightly.)

---

We should have a way to link existing annotations together. (Probably with annotations.) Also, we should offer a way to create a new annotations, end immediately link it to an existing annotation, in one, easy step.

The basic feature (linking two existing annotations) could work like this:

• The user opens an annotation. (Either in the sidebar, or on it's standalone page.)
• User activates the right function. (Which should be right next to "reply" button.)
• A new dialog box opens, where the user can search for the annotation to link. The dialog should have a text entry box for queries, and a result list, updated as he types. (He could also input the ID of the wanted annotation, if he knows it.) We should be able to search for text, author, tags, etc - as usual in both page search, and at the stream interface. In the same dialog, the user could also specify the semantic relation between the two annotations, and any other meta-data necessary.
• When the user has found the other annotation(s), and has configured the wanted meta-data, he click the "link" button, and then the two annotations are linked. (Which on the data layer, probably means that a special, semantic annotation is created.)
• Later, when showing the annotations, the linked partners should be shown, along with the replies. (We need to think hard about this part, especially about the possible different views. Obviously, the "web-of-ideas" thing (#386) comes to mind.)

The extended feature (creating new annotations, and linking them to existing ones in the same step) could work like this:

• User reads the sentence in the new document
• User remembers something relevant, which is already an annotation
• User selects the segment in with the mouse. The adder button, and the new "link" button show up.
• User activates the "link" button, which opens a dialog, similar to the one described above.
• User configures the parameters (like above), and creates "create and link"
• Two things happen:
  • A new highlight is created, from the selected text. (Or full-fledged annotation, if the user has decided to add some body, too.)
  • The new highlight/annotation is connected to the existing one. (Same as in the basic feature.)

Of course, this should not be restricted to text annotations. We could also select an area of an image, activate the right function, and create a new highlight that is linked to an existing text annotation.

[dwhly]

@csillag, this feels very similar to: https://github.com/hypothesis/hypothesis.github.io/issues/15 I think you are describing the workflow for creating a single annotation w/ multiple targets. But you might also be suggesting something slight different.

We are shifting user-facing enhancements to this issue tracker. So either I should close the above issue and merge here, or I should close it and create a new one here.

Are you describing "Multi-target annotations"?

[csillag]

The feature described here is not about multi-target annotations, but about (until finding a better name) meta-annotations, for which the targets are not (segments of) usual documents, but other, already existing annotations.

So instead of allowing an annotation to target selections from several different documents (as described in https://github.com/hypothesis/h/issues/1056), I described a way to add a new layer of annotations, annotating (and linking) the existing, single-target annotations.

This is an alternate approach; both can do (roughly) the same. We should compare them, and collect the advantages and drawbacks of both approaches.

[csillag]

So, the main difference between the two described approaches boils down to this question:

A. Should we do the linking on the level of the "basic" annotations (by adding a new targets to them), or B. Should we create a new, meta-annotation, which describes the link between the two annotations?

I am voting for option B). (Reasoning coming later.)

[tilgovi]

The approach really depends on who's doing the work and what the authoring environment looks like, what user intention it aims to capture, etc.

For instance, I can imagine a simple inverted-linked-list style working fine when there is one author. An annotation has a target which points to the previous annotation. If the trail is A->B->... then viewing annotation A would cause the annotator to look up annotations on A, which would find B.

This whole discussion should really focus on UX and not the underlying technology.

Dan brings up multi-target annotations because some narrative construction might work well as a single annotation with many targets and bodies. Questions open then about assembling the order and flow if they are to be interleaved (as in a Storify, for instance).

I would lean toward avoiding trying to style an annotation so that its bodies and targets are interleaved and instead consider it one body which embeds the targets.

Without a body of some sort, either introducing the narrative or filling in space between the linked annotations, then the resulting document isn't really an annotation because it doesn't have a body. We need to be careful not to imagine everything as annotation.

[tilgovi]

Oops.