Fractals and Semantics

[orion]

Are fractals irrelevant to the nature of semantics?

Think again. This may impact the way we may need to look at the new generation of smart search engines, IR and simbiotic systems, especially now that university, government and private research labs are on the quest for understanding the clear connection of how living neurological networks influence the nature of thoughts.

I have place few references to work along these lines here:


Writing, Grammar, Summarization, Semantics, Knowledge and Fractals (http://www.miislita.com/fractals/grammar-semantics-fractals.html)

Are we ready for bio-searches or the next generation of search engines?

Orion

[orion]

That fractals are relevant to semantics is something well documented in the scientific literature on AI and IR. This is related with scaled patterns found in/through

1. neurological topology
2. grammar
3. languages
4. writing
5. semantic spaces/clustering studies
6. search engine algorithms.

On a kind of related note, in the recent SearchDays article by our very own Chris Sherman, writes

"Writing for a computer that's looking for patterns that can be detected by an algorithm, while simultaneously crafting the same language for a human being who's looking for meaningful information is not an easy task."

http://searchenginewatch.com/searchday/article.php/3492116

Chris is right on target when he talk about patterns. Once we think in terms of patterns, things are a lot obvious. It is time to start thinking about patterns within patterns when optimizing write-ups for both human and computer abstractors.

Orion

[xan]

oh dear :)

[randfish]

Writing for a computer that's looking for patterns that can be detected by an algorithm, while simultaneously crafting the same language for a human being who's looking for meaningful information is not an easy task.
Orion - Is not the purpose of search engines to emulate a human being looking for meaningful information? Wouldn't a writer then have two goals, but the same criteria for each?

I believe that the future of search engine algorithms is to emulate as closely as possible a human being and attempt to find value the same way that humans do. This suggests to me that at some point, successfully writing for humans and search engines will be one and the same thing.

Have you found specific patterns on the web that humans in particular gravitate to? Are these the same ones that search engines are pointing to? Obviously, one goal of IR would be to fuse the two... right?

[orion]

Orion - Is not the purpose of search engines to emulate a human being looking for meaningful information? Wouldn't a writer then have two goals, but the same criteria for each?

I believe that the future of search engine algorithms is to emulate as closely as possible a human being and attempt to find value the same way that humans do. This suggests to me that at some point, successfully writing for humans and search engines will be one and the same thing.



Excellent questions, Rand.

There are many different specialities (neurotopology, cognitive, semantics, linguistics, applied mathematicians, 2G-Knowledge,simbiotic, etc) moving to a common goal: to artificially replicate how the human brain and mind work, process information or generate thoughts. A lot of advances has been conducted by some of my peers and others.

Mankind is getting closer at how the brain and the mind and the several processes (two being the meaning and the "self") work or can be replicated (or at least mimicked). A clear connection between the way neurons are connected forming self-similar patterns and thoughts has been already established.

I have a good collection of such type of research references, if anyone is interested. So when some claims that fractals are not relevant to the nature of semantics I see that as mere non sense.

My take based on such work is that new artificial abstractors are at the horizons of searches. If we start learning in terms of patterns, we can "troubleshoot" a lot better (read here optimize) content that would appeal to both humans and these new machines.


Have you found specific patterns on the web that humans in particular gravitate to? Are these the same ones that search engines are pointing to? Obviously, one goal of IR would be to fuse the two... right?

Please define which patterns you have in mind.


Orion

[randfish]

Let us say, for example, we were to analyze the pattern among a site's on-page text. Would we find certain manners of phrasing, sentence construction, term variety, etc. that would be more "pleasing" to a human reader?

If we were to take that pattern analysis and look at the top 100 or 200 links to a website in Yahoo's Linkdomain results, would we again be able to find certain queues that could tip us off as to the value of those links?

I'm guessing that text analysis and link analysis are the two primary areas we could concern ourselves with as SEOs. Site structure would be the 3rd, but I'm not sure how to phrase that question :)

[orion]

Let us say, for example, we were to analyze the pattern among a site's on-page text. Would we find certain manners of phrasing, sentence construction, term variety, etc. that would be more "pleasing" to a human reader?


Absolutely. Mandelbrot's studies of lexicographic trees explain this and why for instance some specific way of writings (e.g., poems, novels, essays, etc) are "pleasing". (Mandelbrot, B. The Fractal Geometry of Nature, Chapter 38.) Thanks, God, long time ago I have and read his masterpiece.

You may also want to check some of the references given in this essay http://www.lynellen.com/write/restopic.html

For scientific papers, I can provide you with specific references.


If we were to take that pattern analysis and look at the top 100 or 200 links to a website in Yahoo's Linkdomain results, would we again be able to find certain queues that could tip us off as to the value of those links?


That's hard to tell since you are talking now about the Link Contextuality Hypothesis. By just looking at Yahoo's Linkdomain results, one cannot conclude the content of those links lead to are relevant enough to extract a pattern.

Orion

[orion]

I'm off and away from the Internet until Monday. Have all a great, long weekend.

Orion

[orion]

Hi, Rand

Here is another excellent research work on fractals and semantics from J. L. Lemke, titled, Hypertext Semantics (http://academic.brooklyn.cuny.edu/education/jlemke/webs/hypertext/tsld001.htm)

Author: J L Lemke
Source: This TextWeb originated as a presentation in PowerPoint at the 1998 International Congress of Systemic and Functional Linguistics in Cardiff. It represents work-in-progress and will be modified and expanded during 1998-99.

This is the html version of an old ppt presentation.

In the section
Scales for Hypertext (http://academic.brooklyn.cuny.edu/education/jlemke/webs/hypertext/tsld006.htm), Lemke writes

"Finally, we expect in systemic linguistics a certain fractal semantics across scales: the same kinds of meaning relations we make between clauses, we should also be able to make between the macropropositions and macroproposals of text-units on larger scales; this turns out to be true, and relations of Expansion are particularly fundamental to many existing hypergenres"

"In text semantics of longer or larger texts, issues of SCALE are crucial; we need to know what kinds of meanings are made and construed at and across various scales of text, from the usual groups and clauses, to clause-complexes, to rhetorical structure units, to genre element units ... and in the case of hypertexts there are many additional saliences of scale, some of which have little noted precursors in ordinary text (e.g. categorial unity hierarchies)"

Enjoy it!

Orion

[randfish]

Orion,

Thanks for the link - it's much easier to read than most of the stuff I've been looking at in this field :) - Very nice!

I have a question though... If linguistics and semantics in written form are fractal, does this neccessarily mean it can be predicted? For example, if a computer were to examine a text document with one sentence removed, could it conceptually construct that sentence or something similiar based on fractals and predictive technology?

[orion]

Orion,

Thanks for the link - it's much easier to read than most of the stuff I've been looking at in this field :) - Very nice!

I have a question though... If linguistics and semantics in written form are fractal, does this neccessarily mean it can be predicted? For example, if a computer were to examine a text document with one sentence removed, could it conceptually construct that sentence or something similiar based on fractals and predictive technology?

The answer is yes. This has a lot to do with grammar construction and iteration. In the next days or weeks (depending on publisher) a paper that shed some light on this will be published.

Orion

[randfish]

Orion -
That's mind-blowingly cool. Totally incredible. If I'm hearing you right, we could show a computer The Great Gatsby, minus the first sentence and it could tell us that "In my younger and more vulnerable years, my father gave me some advice I've been turning over in my head ever since."

Now that would be artificial intelligence!

[orion]

The answer is yes. This has a lot to do with grammar construction and iteration. In the next days or weeks (depending on publisher) a paper that shed some light on this will be published.
Orion

Actually, thanks to Mike Grehan, there is no need for waiting. The article is now available.

In The Keyword Density of Non-Sense (http://www.e-marketing-news.co.uk/Mar05/garcia.html) article I explain that in
Fractals and Sentence Production (http://www.goertzel.org/books/complex/ch9.html) world famous Ben Goertzel sheds some light about the iterative nature of grammar and lexicographical trees. The way we learn to use iterative patterns helps with the prediction of words. Note the verb-noun rule utilized during learning.

As Goertzel explains, "-The reason N V is a more natural combination is because it occurs at an earlier step in the derivation process."

Other iterative rules have been derived and used in the psychology of languages. These patterns have been used to examine learning problems.

Orion