created, $=dv.current().file.ctime & modified, =this.modified
tags:philosophysemanticsessenceattention
Thought
This is some extent, in relation to Bites with the degree of processable information.
A lot of this thought seems to circle around Attention, and the degree we have distilled. Which calls to mind
“Attention without feeling is only a report” Misc vs.
“Attention is all you need”
Why is that transformer models and all that immediately dominates the conversation and immediately renders the search below boring af? All while similarly not really bringing any sense of closeness to the problem, just leveraged by previous solutions. (this comment could also be completely wrong, because I am ignorant.)
How do we effectively summarize a body of text? What is meant by the [essence] of the text? What criteria will place one summarization above another?
ROGUE (Recall-Oriented Understudy for Gisting Evaluation) relies on previously written human reference.
What do we pay attention to? What are we conscious of?
Rather than focusing on what is gained (as in what the summarization leaves intact, which seems to be the focus - the result), what is lost through summarization? All of those particular elements that don’t make the cut. What are they? What is the relationship with what is lost?
NOTE
I might have later hear of this in a podcast, referred to as noise. I wasn’t thinking of it as noise here but want to document it like that. It give an interesting perspective, like tuning (fine-tuning), being the amount of noise you find acceptable. That acceptance doesn’t really change the shape of the noise (the words lost in summarization remain the same), it’s just a way of seeing.
What can we do with those scraps? If we combined them all after a task.
Related
After writing this I discovered the project moDernisT by Ryan MaGuire that called this concept to mind. MP3 compression is lossy. Bits of the song are discarded, filtered out. This project surfaces these missing sounds and plays them back in isolation.
The MPEG-1 or MPEG-2 Layer III standard, more commonly referred to as MP3, has become a nearly ubiquitous digital audio file format. First published in 1993, this codec implements a lossy compression algorithm based on a perceptual model of human hearing. Listening tests, primarily designed by and for western-european men, and using the music they liked, were used to refine the encoder. These tests determined which sounds were perceptually important and which could be erased or altered, ostensibly without being noticed. What are these lost sounds? Are they sounds which human ears can not hear in their original context due to universal perceptual limitations or are they simply encoding detritus? It is commonly accepted that MP3’s create audible artifacts such as pre-echo, but what does the music which this codec deletes sound like? In the work presented here, techniques are considered and developed to recover these lost sounds, the ghosts in the MP3, and reformulate these sounds as art.
Thought
Do I ever encounter a core idea? When I pointlessly read and research, say I am reading wikipedia. This is a summary. Say I read a book? Closer to the core? Perhaps, but still this feels “summary.”
What can I do to fully experience a core idea? To live it? To set aside a picnic on the grass, feel the lemonade pour down my throat, the heat and the presence of a friend.
No, this is “summery”.
Shopping
This is one place where AI has been applied. One example being this new segment of the Amazon shopping experience. When there are sufficient reviews for a product an AI generated section will appear on the page that attempts to distill the text, by linking keywords with references.
Rather than spending the time to read an individual review (most people balk at the sight of paragraphs these days, unfortunately but occasionally with reason), they’ll want the bullet points. I am charitable to reader, in believing that this could be an effective way of sorting through the mass of text and pulling out a specific review to read.
What we find is that in the case of reviews
- Seldom are you unique, not necessarily as a consequence of an identical product being the focus of the review
- As in your review of the vacuums will likely not involve discussion of clouds, or how well it handles deep sea cleanup.
- People want brevity
- Star ratings are built in
- A further distillation, stars themselves or any numerical/categorical rating are point based summarization. A reduction.
- It seems this fundamentally is the method at play here? In that it’s “star ratings” all the way down to construct these review mappings.
Thought
How can you provide a numerical rating that isn’t a reduction?
Semantic Similarity
Abstracts
An abstract is a brief summary of a research article, thesis, review, conference proceeding, or any in-depth analysis of a particular subject and is often used to help the reader quickly ascertain the paper’s purpose
History of
The history of abstracting dates back to the point when it was felt necessary to summarise the content of documents in order to make the information contained in them more accessible. In Mesopotamia during the early second millennium BCE, clay envelopes designed to protect enclosed cuneiform documents from tampering were inscribed either with the full text of the document or a summary. In the Greco-Roman world, many texts were abstracted: summaries of non-fiction works were known as epitomes, and in many cases the only information about works which have not survived to modernity comes from their epitomes which have survived. Similarly, the text of many ancient Greek and Roman plays commenced with a hypothesis which summed up the play’s plot.
Reminds me of early film reviews. We’ve lost some of the original films but reviews of the missing films still exist.
With respect to Science, (which is how I think of it), the use originates in the early 1800s, when the secretary of the Royal Society would record brief summaries of talks into the minutes of the meetings. These were later collected and published.
“Only reading the headline:”
Abstract is often expected to tell a complete story of the paper, as for most readers, abstract is the *only part of the paper that will be read.
TL;DR
Originated in 2002 in usenet newsgroup rec.games.video.nintendo
Transformers
Summarization can be:
- extractive: extract the most relevant information from a document.
- abstractive: generate new text that captures the most relevant information.
Visual
Summarization is not just a quality of text. Visual summarization can occur. Imagine a keyframe based extractive mechanism for real life which would capture a moment in the day as in Automatic Media.
What constitutes the best “1,000 words” a picture is made of?
Trailers
Trailers have an aspect of summary.todo
Youtube
Youtube videos now have AI summarization on the page.
