Forum Posts Feed
Is Semantics search the end of 'information organization'?
< Next Topic | Back to topic list | Previous Topic >
Posted by Chris Thompson
Apr 19, 2018 at 06:27 PM
You’d probably enjoy taking a look at this paper: http://students.ecs.soton.ac.uk/mwra1g13/msc/comp6045/pdfs/Bernstein%20-%20Can%20We%20Talk%20about%20Spatial%20Hypertext.pdf
It’s not very technical, but it’s an interesting exploration of various spatial hypertext idioms from the author of Tinderbox. Tinderbox itself emerged from Bernstein’s graduate work in spatial hypertext; there are tons of really cool old spatial hypertext research systems like VIKI that few people have ever heard of but which introduced a lot of interesting ideas. The list of references in that paper provides a starting point for reading about them, and there is an interesting, opinionated discussion in Bernstein’s “The Tinderbox Way” book.
There’s also a whole body of work on the “semantic web” (basically, hypertext data representation with labelled arrows; think arbitrary concept maps), but much of that seems to have petered out. This Google project uses “semantic” in a different sense, referring to its traditional meaning of language understanding.
—Chris
Dellu wrote:
>
>Paul Korm wrote:
>>You might be interested in some of the research published in this area
>>through the ACM (Association of Computing Machinery), which has a
>number
>>of SIGs specializing in semantic computing, search, hypertext, and
>>related fields.
>
>Thank you. I am totally unaware of the academic literature on this. I am
>checking it out.
Posted by Dellu
Apr 19, 2018 at 07:02 PM
Chris Thompson wrote:
>There’s also a whole body of work on the “semantic web” (basically,
>hypertext data representation with labelled arrows; think arbitrary
>concept maps), but much of that seems to have petered out. This Google
>project uses “semantic” in a different sense, referring to its
>traditional meaning of language understanding.
>
>—Chris
Hypertext is interesting means of organizing information. my understanding of “semantic search” is different from hypertext. For Tinderbox type of system, you need to explicitly assign “tags” to give a meaning to your content. “semantic search” could make the assigning of the tag superfluous because the system already knows conceptual associations (lexical synonymous; , artificially intelligent correlations between concepts) that you don’t need to assign key terms or hyperlinks. To my mind, this latter approach is much more potent. Comparing it with the semantic search implemented in the google system, Tinderbox’s type of mapping feels “manual” organization of information because you have to manually tag the attributes. Probably, I don’t understand hypertext well; specially the ‘arbitrary concept map’ part.
Posted by Dellu
Apr 19, 2018 at 07:26 PM
I put “Norway prosper” into the search engine, some of the results I get are very fascinating.
>In Norway, after the discovery of oil in the North Sea in 1969, economic growth accelerated for the following 25 years, allowing the country to catch-up and then exceed its otherwise highly similar Scandinavian neighbours, Denmark and Sweden, in terms of GDP per capita (Larsen, 2005)
>After Norway became an oil-producing nation and the national income per capita rose above that of Sweden, the self-image of the Norwegians seems to have been strengthened
None of these outputs actually contain the term “prosper” nor any of its close Synonymous.
If I can magically fetch the exact concepts I am looking for like this, why do I need to keep a list of quotations or summaries or the notes from the books in my note writing software?
In this kind of powerful semantic search is in our hands, the organization of knowledge by manually setting maps/tags or hyperlinking concepts is successfully outdated.
Posted by Chris Thompson
Apr 19, 2018 at 07:49 PM
That’s true, the need to manually tag is why the “semantic web” concept ended up petering out, though it spawned piles and piles of research papers on “ontologies” and “triplet stores” and a lot of other things. Intelligent untagged search (“semantic search”) just ends up being more powerful because no one has the time to tag the universe or apply all these standardized ontologies.
But as a researcher/notetaker, the task is a little more focused. There’s no shortcut to actually taking notes based on your own understanding in order to develop your personal understanding of a domain. It’s why having three dozen research papers on your hard drive ends up not being all that enlightening—even if you had a perfect semantic search engine for those papers—while having little personal write-ups/summaries, extracted personal highlights, and being able to sketch out interesting relationships you’ve noticed between concepts in those papers tends to lead to more complex understanding and the sort of insights you can develop to write novel papers/books. It’s this latter thing that hypertext systems like ConnectedText and spatial hypertext systems like Tinderbox are designed to help with.
—Chris
Dellu wrote:
>Hypertext is interesting means of organizing information. my
>understanding of “semantic search” is different from hypertext. For
>Tinderbox type of system, you need to explicitly assign “tags” to give a
>meaning to your content. “semantic search” could make the assigning of
>the tag superfluous because the system already knows conceptual
>associations (lexical synonymous; , artificially intelligent
>correlations between concepts) that you don’t need to assign key terms
>or hyperlinks. To my mind, this latter approach is much more potent.
Posted by Dellu
Apr 19, 2018 at 08:18 PM
Chris Thompson wrote:
>But as a researcher/notetaker, the task is a little more focused.
>There’s no shortcut to actually taking notes based on your own
>understanding in order to develop your personal understanding of a
>domain. It’s why having three dozen research papers on your hard drive
>ends up not being all that enlightening—even if you had a perfect
>semantic search engine for those papers—while having little personal
>write-ups/summaries, extracted personal highlights, and being able to
>sketch out interesting relationships you’ve noticed between concepts in
>those papers tends to lead to more complex understanding and the sort of
>insights you can develop to write novel papers/books. It’s this latter
>thing that hypertext systems like ConnectedText and spatial hypertext
>systems like Tinderbox are designed to help with.
>
>—Chris
I fully agree. I just considered these points more of part of knowledge generation (rather than knowledge organization). I indeed need to structure my own take/points, deductions and extensions, of the ideas(concepts) in my notes. that is a way to construct and generate new knowledge. It is also possible that a reader constructs conceptual links or deductions that no artificial intelligence could emulate.