Reconstructing existing articles as Narratives

[csillag]

Migrating this from https://github.com/hypothesis/h/issues/767

---

One of the usage modes I would love to see evolving is reconstructing existing documents into Hypothes.is narratives (see #14), and thereby making them easily accessible for discovery and further discussion inside our system.

Let me explain.

Level 0 is the current situation. Level 1 is the introduction of narratives (#14); starting from that point, we can start to build new documents that way. Level 2 would be supporting the reconstruction of legacy documents (which were constructed with the current process) using the new process. This is what the current issue is about.

To do this, we will need some special tooling. The work-flow with such a tool could look something like this

• Open the tool. (Which should be a full-screen app.)
• Create a new narrative with importing an existing document. At the start of the process, the whole text of the original document (including the quotes) would appear as comments in the narrative.
• Supply some meta-data. Who is the author of the document we are working with? What are the source documents which will be referenced?
• When all the meta-data is in place, the user should start to go over the narrative. This could be a semi-automatic process. Something like this:
• The user identifies a part of the text which is a quote. He selects it, and presses a button. The button is very, very smart, and does bunch of things:
  • It checks all the existing annotations and highlights (stored for the defined source documents) for a match. If it finds a match, then that means that we have found our quote. We can replace the quote with an embedded highlight/annotation.
  • If it finds no matching annotation/highlight, then we can do a (fuzzy) search over the texts of the referenced source documents, and see if we can find the quote. If we can, then we automatically register a highlight for the quote, and replace the quote in the narrative with the embedded highlight.
  • If we can neither find, nor automatically create the referenced highlight, then we ask the user what to do. Some of options which should be offered:
  • Is this a quote from somewhere else? Give me the URL for the document to scan.
  • Manually enter ID of a highlight (which should then be verified by the system).
• Repeating this process, all the quotes that were copied into the original article should be replaced by embedded highlights. We should have dedicated buttons on the UI for doing this quickly. (Find something that looks like a quote, fix it, move to next quote, mark something as a quote, etc.)
• When the document has bean "cleaned" (that is, all the copied quotes have been replaced by embedded highlights), we can publish the new narrative.

The re-constructed narrative is an enhanced, more interactive form of the original document, since all the quotes are clickable, and can be directly looked up in the source documents; they can be replied to, etc. (Furthermore, since the highlights in the source documents also get replies with a link to the new narrative, the connection will be automatically bi-directional.)

Since narratives would be so nice, I would also consider automatically offering the users to read the reconstructed narrative instead, whenever we deploy H over a document which we have a reconstructed representation for.

[tilgovi]

How is this not a duplicate of #14?

[csillag]

As described in the opening post:

• Level 1 is the introduction of narratives (#14); starting from that point, we can start to build new documents that way.
• Level 2 would be supporting the reconstruction of legacy documents (which were constructed with the current process) using the new process. This is what the current issue is about.

[tilgovi]

You're going to have to explain. I cannot see how this is different from #14.

[csillag]

You're going to have to explain. I cannot see how this is different from #14 https://github.com/hypothesis/hypothesis.github.io/issues/14.

14 is about building support for telling your story by building a

narrative.

This issue is about building support for taking an existing article, and reconstructing it as a narrative, using a semi-automated process.

This requires a different UI than free-form narrative building, because here the goal is to arrive at a narrative which is equivalent to a pre-existing document, by converting the legacy article into a narrative piece by piece.

So the difference is like (for example)

• Level 1: An editor for writing stories
• Level 2: A computer-aided-translation suite for automatically rewriting already existing stories in a new language.

Is it clear now?

[tilgovi]

No. They seem like the same workflow but with different content. Why does it matter whether one is rehashing an existing narrative or making a new one? In either case they're writing documents which embed quotes. Which is why I also feel like this is duplicated by #15 as well. I can't tell all these apart. They seem like not different features.

[csillag]

No. They seem like the same workflow but with different content. Why does it matter whether one is rehashing an existing narrative or making a new one?

Because there should be some features specifically for augmenting the rehashing of an old one. If you read the description, you will see what those are. (Like automatic recognition of quotes, etc.)

In either case they're writing documents which embed quotes.

No. In the default case (#14), you are building a free-form data structure from existing quotes + commentary from the ground up.

In this case, you are working with an already existing text document, and you are slowly transforming it into a data structure made up of quotes and commentary, without actually loosing the content, or adding anything new.

---

Let's talk about the imagined UIs for each of these features later, and you will see the difference.

[ajpeddakotla]

Closing in favor of: https://github.com/hypothesis/h/issues/385