Zettelkasten Forum


Tagging individual sentences (rather than the entire note)

I've recently become involved in some qualitative research that involves using the program NVivo to analyse interview transcripts. Basically, the interview transcript is imported into NVivo, then you go read each sentence and attached thematic tag(s).

I think this would be an interesting way to organise research and notes, with the understanding that the tagging system would evolve over time.

For example,

If you are working on a specific article, you could browse through previous semi-related tags and tag sentences that might be relevant to that article.

Or if you are working on a longer project, tags could start general but then over time be refined and become more specific as you work out the structure of the project.

The only way I can think to implement something like this in existing note taking apps is to have one sentence per note, then as many required tags per note, but this seems unmanageable.

I'm curious what other think of this idea?

Comments

  • After a bit of searching, this seems to be closest to what I'm looking for https://app.taguette.org/

  • Can you explain what that app does or share a picture of how you use it? Tbh I can't quite imagine what's going to happen, looking at the website alone.

    Some apps support #hashtags so you could annotate sentences in-line as well. I find this a bit unwieldy. But then again I'm not analysing interview transcripts, so it'd be interesting to see how you incorporate these.

    Author at Zettelkasten.de • https://christiantietze.de/

  • JonJon
    edited January 2022

    For collaborative/academic reasons I've had to use NVivo for the interview transcript analysis, so I've not actually used Taguette for any projects.

    As a quick demo,

    Imagine two people responding to an interview question asking about their views on the potential impact of artificial intelligence.

    You upload the files and you then can go through and highlight/tag each sentence.



    The idea is that the research team will go through the interviews and iteratively develop the tagging scheme, then continue the interviews until "thematic saturation" is achieved. This could be defined as 5 interviews where there have not been any new themes and all expressed ideas have been able to be captured by your tagging scheme.

    You can then report on the themes that have come up in the interviews.

    I think it is (maybe superficially) similar to the Zettelkasten in some ways, in that there is an eventual emergence of themes over time. Maybe the tags act as kind of a well maintained structured note in a Zettelkasten.

    Tagging sentence by sentence is probably too burdensome and granular though. And there is an overhead of maintaining the tags over time. Though this could be an interesting way to continually synthesise and reinterpret your collection of notes.

    (Note: I'm really inexperienced with qualitative research, so I may be misinterpreting things)

  • Though, it is a bit outdated: This post might contain something for you.

    https://zettelkasten.de/posts/intratextual-tags/

    I am a Zettler

  • The only way I can think to implement something like this in existing note taking apps is to have one sentence per note, then as many required tags per note, but this seems unmanageable.

    I'm not dealing with qualitative research myself. That said, I don't think that having one sentence per note is unmanageable. On the contrary, I think that having many individual notes can be very powerful & flexible. And the usual ways to structure your notes (like keywords/tags/labels, links, structure notes, ratings, citekeys/references, etc) can help you to search, filter & arrange them.

    To illustrate, here's a made-up example using my own (still unreleased) app. But this should be possible using any typical note-taking or Zettelkasten-style app that allows to structure & filter again your notes.

    Having your tagged highlights available as individual plaintext files has the usual benefits, esp. the possibility to further work on your data using third-party tools.

  • Argdown is an existing Markdown-compatible lightweight markup language and toolset that allows tagging individual sentences/propositions, specifying other aspects of the argument structure, and visualizing it all as an argument map. It is part of the field of argument technology along with a lot of other related kinds of software.

    The more general issue here is the level of granularity of the knowledge elements in a personal knowledge base. Once you start wanting to tag individual sentences, then you may want to link between individual sentences too, and you may want both the sentences and the links to be typed, and pretty soon you have a highly granular knowledge ontology, far beyond an ordinary zettelkasten. For example, Argument Interchange Format is an argument ontology that is like a more complex version of what you find in Argdown.

    I am also reminded of the Text Encoding Initiative (TEI), a community of humanities scholars that developed guidelines for creating highly granular XML schemas, like "HTML on steroids". It's a good example of how crazy you can get with the granularity. (Pandoc can convert files to the "TEI Simple" schema.)

    By the way, I have also used the writing application Scrivener for this purpose. A Scrivener project can be configured so that there is one sentence/proposition per document (note), so each sentence/proposition can have a variety of metadata and can be individually linked to. This is manageable because Scrivener easily allows you view an arbitrarily large number of documents together in one scrolling window. The Scrivener project can then be compiled to other formats.

  • @Sascha Thank you for the article link. I think one difference between the intratextual tags shown here and individual sentence (or paragraph?) tagging, is that the search for the tag will return the whole note rather than just the individual tagged section.

    @msteffens Your app and example looks very interesting! I guess having one sentence per note could be manageable if they were able to be well displayed. For example, perhaps the back-end of an app may save each individual sentence as its own unique entry, but on the front-end be able to display them to the user in the paragraphs that they were entered.

    @Andy Thank you, I had never heard of Argdown, the field of argument technology, or the Text Encoding Initiative. I agree that the general issue is one of granularity. It's something I've struggled with for a while with note taking. I had not thought of using Scrivener for this purpose. I can imagine it could be very beneficial to have individual sentences entered and tagable as their own notes, but then viewable as a continuous whole. From memory scrivener also lets you re-arrange the order of the files, which could help with producing an article/piece of writing.

  • @Jon Perhaps Logseq might be an app for you. It follows a quite granular approach

    I am a Zettler

  • @Jon said:.

    The only way I can think to implement something like this in existing note taking apps is to have one sentence per note, then as many required tags per note, but this seems unmanageable.

    I'm curious what other think of this idea?

    Solution 1:

    In Obsidian, and Craft, , you can:

    • Append "tag notes" to a sentence you need to review, or "any verb" later
    • Open the "tag note"
    • Look at the backlinks at the bottom of the tag note, where you'll find a list of notes which had that tag note added, and, for each note, the sentences with that tag note.

    Obs.: you can have as many sentences as you like per note.

    Solution 2:

    In Obsidian,

    • Append #tags to sentences
    • Use the Dataview plugin to build the queries that are relevant to your use case

    Obs.: you can have as many sentences as you like per note.

  • @gdigesu: Those are great ideas about how to tag individual sentences in Obsidian, which is a very promising app for the qualitative-research use case that @Jon mentioned above! In fact, Ryan J. A. Murphy wrote a blog post seven months ago about "An Integrated Qualitative Analysis Environment with Obsidian".

    But it gets even better: In Obsidian, you can link blocks or transclude blocks from one note into another note, either with automatic block IDs or with a custom/manual ID at the end of the block. So depending on how you are going to process the data, you could just put each sentence (or other semantic unit) in its own block, that is, a separate paragraph or list item, instead of in a separate note. So each interview or other source text could stay in one file, split into blocks tagged as @gdigesu described above, queried with the Dataview plugin, and transcluded in other notes.

    However, Ryan J. A. Murphy's blog post shows that there are advantages of keeping each sentence or semantic unit (or "data point" as Murphy calls it) in its own note and adding each tag (or "code" in qualitative analysis) as an internal link to a "tag note". The following advantages of this system are listed in Murphy's blog post:

    • Codes are suggested when you start a new internal link [[, so that we can choose from the options instead of trying to recall them.
    • The individual responses will be structurally related to the codes via backlinks (and visually linked to the codes via the graph view).
    • Editing a code by renaming it will propagate the change to that code throughout the data.
    • Finally, it makes codes themselves an object we can add information to. You can open a code-as-link as a note and describe it, embed exemplary instances of the coding from the responses, and so on.

    Particularly cool in Murphy's list of advantages of one data point per note is the ability to see the links between data points and codes in Obsidian's graph view. If you just tagged multiple sentences in a single note, Obsidian's graph view would be much less informative.

    By the way, it also occurred to me recently that this topic is somewhat related to semantic line breaks (also called semantic linefeeds). The exact nature of the relation is left as a fun exercise for the reader.

  • @Andy Very good remarks.

    On the course of management consulting engagements, we do lots of both quantitative and qualitative analysis.

    The qualitative analysis, back in the 1980's, was done with physical cut (with scissors ...) and paste (glue ...) 😆, Xerox copies, and a typist heading the table.

    Then, in Wordstar, or Word 🤓, which was a phenomenal improvement!

    Nowadays, it is a walk in the park 😎.

Sign In or Register to comment.