Nodal Point of Thought 2023-08-10-2
gpt.iconSpecific use of Kozaneba: I would like specific information on what kind of tool Kozaneba is and how it is used. Further explanation of how it is used to organize time series information and what difficulties were encountered would be needed.
Vector search.
Q: Is Kozaneba a tool that is both time-oriented and topic-oriented?
A: I wonder. First of all, I think that when the sentences in a book are chopped up and arranged, it is a temporal scheme based on the order in which they appear in the book. Then, during the phase of shaping the text, I think the scheme changes to one that reflects the syntactic structure of the sentence.
That's all I'm going to use in the process of making this presentation, so I'm going to read the explanations in order.
There are some "chunks" that are lumped together, such as twin dangers and side definitions.
Kozaneba is a "tool that can do both" after all, so it depends on how users feel about using it.
Q: Is it ideal to be able to move freely between temporal and contextual schemes?
A: I think it would be ideal for me to be able to come and go.
When I'm organizing, I want to organize on the topic axis, but when I don't understand a fragment by itself, I want to shuffle around and see the time line where it came from, and then I see, and then I want to go back to the topic axis and organize again.
I have been thinking that it would be nice if Kozaneba had such a function, but now "position" is a two-dimensional vector, which is just a higher dimensional vector mapped to two dimensions, with a time axis behind it. But that's ok.
Q: Essentially what we need is a dependency of information, and I think temporal relationships are a substitute for that
A: When expressed as a sentence, dependencies naturally tend to be in close proximity in a sentence
So when you read it, you need to read it once you have preserved the information about the time frame, and then afterwards you need to read it and wonder what the structure would have been like in the author's mind.
nishio.iconI didn't realize how important it was to "draw the line" when I created Kozaneba.
The best part of creating Kozaneba may have been the verbalization of "a line is necessary.
human.iconIndeed, it's great fun to see that kind of trial-and-error process reveal the necessary elements!
1
It is difficult to create a new structure when there is an existing structure "when expressing a relationship in proximity".
This is because "moving" is immediately "destroying existing structures," and "existing structures that are being broken" and "new structures that are being created" are expressed as a whole without distinction.
In the process of using Kozaneba, I thought, "Wouldn't it be better as a sample if Kozaneba's release notes themselves were organized in Kozaneba?" I tried it, and even with my ability, it was impossible to "create a new structure from the data once structured in chronological order without shuffling the structure.
After thinking about why this was the case, we decided that it was wrong that the only way to express relationships was through proximity, so we added a line-drawing function.
---
This page is auto-translated from /nishio/思考の結節点2023-08-10-2 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.