⏾ Concept Map

All about encounters

|↑| Records | All about encounters

This page gives a brief description of the learning method which builds the map as it goes.

When reading something and trying to understand it, I highlight the bits that I don’t “know”. The question “What is that?” leaves behind a highlight each time it pops up in my head. After finishing the material completely, I gather the highlighted terms and try to make sense of them; they eventually become the concepts that enter the site.

By extension, the same method can be used to learn from any medium where something analogous to a highlight may be recorded. Due to their ubiquity and ease of manipulation, I find that screenshots are a good choice for non-text-based mediums, and I use them in place of highlights for most anything that isn’t a structured document.

Whether a highlight or a screenshot, the entries titled encounter capture this initial context which made me think to capture an idea. What follows is a list of mediums and the tools and methods I use for each, plus some details like efficient timing for processing.

  • For academic literature like papers and textbooks…
    • Modality: PDF in Zotero. Could use any device, but I like it on a tablet.
    • Form: The surrounding page (cut the PDF).
    • Extraction: Custom script using pymupdf.
    • Citation: Point to the source’s own node in-map.
    • Timing: …
    • Comment: Zotero retains the original document with highlights as a canon.
  • For one-off documents like lecture slides…
    • Modality: PDF in Zotero. Could use any device, but I like it on a tablet.
    • Form: The surrounding page (cut the PDF).
    • Extraction: …
    • Citation: …
    • Timing: …
    • Comment: Zotero retains the original document with highlights as a canon.
  • For plain literature like novels…
    • Modality: EPUB in KOReader. Could use any device, but I like it on an e-reader.
    • Form: The surrounding paragraph as text.
    • Extraction: Custom script using slpp and lxml.
    • Citation: Point to the source’s own node in-map.
    • Timing: After finishing the book.
    • Comment: KOReader retains the original document with highlights as a canon.
  • For websites…
    • Modality: No preference. Any modern web browser will do.
    • Form: The surrounding paragraph as text.
    • Extraction: Copy it then and paste it into the staging area. Strip the formatting unless doing so is destructive.
    • Citation: Point to the URL. It may indeed break in the future, but that’s okay; the URL and blurb themselves are sufficient. If it’s a newspaper, point to the source’s own node in-map and the URL.
    • Timing: …
    • Comment:
  • For chatbots…
    • Modality: No preference. The interface isn’t important, but the model is; I like Claude.
    • Form: Text as given by the provided copy-to-clipboard button.
    • Extraction: Copy it then and paste it into the staging area. Strip the formatting unless doing so is destructive.
    • Citation: Point to the source’s own node in-map.
    • Timing: …
    • Comment: Specify the model family rather than the product line — they update default models and move things around so frequently that it’s easy to swap one for another without realizing. The increased specificity doesn’t add much anyway. (“Claude”, not “Claude Sonnet”.)
  • For moving media like video and video games…
    • Modality: No preference. Anything that can take a screenshot.
    • Form: A subtitled screenshot in a standard resolution.
    • Extraction: …
    • Citation: Point to the source’s own node in-map, or the URL for things like YouTube videos.
    • Timing: After finishing the material.
    • Comment:
  • For terminal dialogue like diagnostic messages from a software package…
    • Modality: No preference.
    • Form: The relevant lines as text.
    • Extraction: Copy it then and paste it into the staging area. Strip the formatting unless doing so is destructive.
    • Citation: Point to the source’s own node in-map.
    • Timing: …
    • Comment:

Annotation Keywords

A small number of keywords are used in footnotes or comments to flag unique items.

  • ”?”: Low-confidence identifications.
  • ”???”: Skipped items.
  • “compound”: A single highlight mapping to more than one concept.
  • “marginal”: Likely duplicates or low-value items. These are usually deleted before they make it to the site.
  • “near”: Near-misses/matches in identification, like by using a close idea instead, or a synonym for vocabulary items.