Re: For each bubble, a container for KR terms

Tyson,  thanks for your inquiry

Your question shows understanding the big picture and a propensity of
integral, systems thinking
and functional decomposition, which I can relate to

Your approach to distinguishing different levels of analysis can be useful,

This is also my first post to the S AGENT community, so I ll try to make it
sound like a position statement
I am jumping at the chance to evangelize about KR, once more

For many of us, This has been/still is a steep learning curve

At the top of the hierarchy is LOGOS   *logic
  --------Lookup different types of logic/logical formalisms/logical
languages

-----------Different forms of logic may be suited for different kinds of
reasoning mechanisms and different languages may be suited for
better representing logic and reasoning, depending on the use case, the
environment , the platform etc

In the last decade or so, we have seen a wave of machine learning AI
driving scientific and technological progress
I found the notion of   'fragmented logic'  in AI systems problematic
i discuss the concern here in our first report
https://www.w3.org/community/aikr/wiki/images/7/7b/AI_KR_FIRST_REPORT_PUBLISHED_VERSION.pdf

More recently we are focusing on LLM *forget the rest of AI for a moment
(too big a topic)

Knowledge Representation *in itself a vast field that developed
alongside computation in the last century (among other fields to which KR
is relevant)

KR that uses NL

can help humans stay in the loop, by supporting natural
language understanding *NLU
thus enabling experts with multidisciplinary backgrounds to share knowledge

Now we are doing LLM evaluations that require domain experts to validate
the outcomes of LLMs (across a number of fields)
Fools do not speak machine language, nor use maths to communicate

*I*intelligent reasoning and human understanding of what constitutes
intelligent reasoning,  are represented by logic, which is expressed with
either human natural languages and/or with formal languages, Mathematics is
one such language. Description Logic DL is another

While structural engineering uses calculus , architects use drawings to
convey shapes, spatial concepts, where windows and doors are place, their
size etc.  KR embraces all of these formalisms, the designer.engineer
selects the most appropriate ones according to their goals, methods,
resources.

Here in this CG *AI KR CG we have aimed so far to chart the field of
knowledge representation so that when we try
to understand LLM outputs we have a compass for orientation rooted in Logic
which can be represented and interacted with using  Natural Language

Referencing some very old conversations if you can find a way of querying
Ontology archives before it migrated to Google Group
http://ontolog.cim3.net/forum/ontolog-forum/2007-03/msg00150.html

A lot of interesting conversations there about KR *where I learned the ropes

In the link below some notes showing a graphical representation of how a
human problem solver would represent knowledge about the engine
not starting
https://docs.google.com/document/d/1dKVibyqHJ3Qoy7WrqPUYi6iwZN94T5puJrgKEJFYV1I/edit?usp=sharing

I am surprised how good translation of notations works across formalism
These days we can translate any formalism into any other formalism
including hieroglyphs!!

Is mathematics a KR language ? Yes!
Is it a natural language that humans and LLMs communicate with?  No!


So much more to say!  So now we are focusing on publishing a few
vocabularies and schemas that represent
the KR domain,  to help us navigate the evolving field of AI and support
our LLM  evaluations

I look forward to be learning more about the S AGENT COMM is dong and how
can KR contribute to it
and if ou have anthing relevant to contribute to our next report to be
published soon *a shared concern, or a shared solution
please lets crosslink

Best regards

Paola Di Maio


On Sat, Nov 22, 2025 at 12:25 PM 陳信屹 <tyson@slashlife.ai> wrote:

> Thanks again to all of you for the very insightful sharing.
>
> As someone who is newly entering the KR domain, I have been trying to
> understand the overall landscape by organizing the concepts using the
> following layered architecture.
>
> ```
> [L4] Semantic Execution Layer (our work: semantic agent communication CG)
>     └─ OS-level execution, agent sandboxing, DID/VC identities,
>        semantic ledger, cross-agent message schemas
>
> [L3] Semantic Behavior Layer (our work: semantic agent communication CG)
>     └─ Contracts, capabilities, delegation structures,
>        execution-context models
>
> [L2] Semantic Positioning (Milton)
>     └─ Conceptual placement of KR between
>        language-sensitive context and application-level behavior
>
> [L1] Semantic Materials (Paola)
>     └─ Vocabulary sets, definitions, classifications,
>        semantic maps, knowledge-representation corpora
>
> [L0] Semantic Atoms (Daniel)
>     └─ Character-level units combining form, meaning,
>        procedural RPN programs, and embedding space
> ```
>
> My intention is simply to understand how different lines of work relate
> along the spectrum from semantic materials and definitions (L0–L1), to
> conceptual positioning (L2), to behavior-level structures (L3), and finally
> to execution frameworks (L4) .
>
> If this mapping misrepresents anything, I would be grateful for your
> guidance.
> Thank you again for the discussion and for helping me understand the field
> more clearly.
> Paola Di Maio <paoladimaio10@gmail.com> 於 2025年11月20日 週四 上午10:37寫道：
>
>>
>> Milton and others
>> I am glad there is interest in KR and the work being done by this CG,
>>
>> Please feel free to share your work, you do not need anybody's permission
>> to draw diagrams, write papers or present your
>> evaluations *provided these are related to the work being done
>>
>> It is quite difficult to get people to read our work, this is why we must
>> validate our own
>> work externally and try to publish it *better still if it has some
>> evaluation and validation built in
>>
>> I am now training  agents *as per the talk on KRL and have become very
>> absorbed, so I  need to focus
>> my attention where it is needed
>>
>> Next steps for me
>> - try to wrap up some vocabs (possibly a subset) the open them up for
>> evaluation and consultation
>> - find out if and how to move forward toward a specification
>> - training Agents to  support this work.
>>
>> CG participants with the required competences and resources please get in
>> touch to help with the plan
>>  we may cook something together
>> and share the results with this list
>>
>> Look forward to be reading your papers or listening to your talks on KR,,
>> Milton and everyone, whether it is literature reviews, state of the art
>> reports or new advances
>>
>> Best
>> Paola
>>
>>
>> On Thu, Nov 20, 2025 at 2:25 AM Milton Ponson <rwiciamsd@gmail.com>
>> wrote:
>>
>>> Maybe it is necessary to explain my point of view in plain language.
>>>
>>> Knowledge representation is "the man in the middle" between knowledge
>>> and application.
>>>
>>> Both ends define the contexts for the man in the middle, and both ends
>>> are determined by specific domains of knowledge or application.
>>>
>>> We can create general frameworks on both sides of the man in the middle.
>>>
>>> On the left hand side natural language context sensitive and right hand
>>> side computationally application specific.
>>>
>>> Maybe we should redraw the bubbles graph to reflect this.
>>>
>>> Milton Ponson
>>> Rainbow Warriors Core Foundation
>>> CIAMSD Institute-ICT4D Program
>>> +2977459312
>>> PO Box 1154, Oranjestad
>>> Aruba, Dutch Caribbean
>>>
>>> On Wed, Nov 19, 2025, 12:12 Paola Di Maio <paola.dimaio@gmail.com>
>>> wrote:
>>>
>>>>
>>>> Just to remind newcomers, and participants who may have not followed
>>>> the work being done in detail
>>>> or may have lost sight of the mission *admittedly we have had wide
>>>> ranging conversations on this list
>>>>
>>>>  that over the years, hundreds of terms representing the KR domain have
>>>> been
>>>> collected, refined, and evaluated for inclusion from the corpus
>>>>
>>>> Not just academic literature, but lecture notes, from use cases and
>>>> technical documents in use.
>>>> Some of these vocabs are being peer reviewed, but still far from being
>>>> final!
>>>>
>>>> So for each bubble in the diagram  *note I have produced a new version
>>>> with links to the vocabulaires, but at this moment these are not shared
>>>> resources yet
>>>>
>>>> [image: AI KR VOCABS NOV 2025(1).jpg]
>>>>
>>>>
>>>>  providing an overview of the conceptual spaces in KR, there is
>>>> vocabulary capturing hundreds of terms and concepts extracted from the
>>>> KR corpus.   Would not say this is quite 'core vocabulary for KR yet *
>>>> because it needs refinement
>>>> We have over 1400 unique terms from one category alone
>>>>
>>>> I am confident that it is the most extensive map of the KR domain  *and
>>>> it is becoming difficult to handle
>>>>
>>>> For each vocabulary, now I have different versions. Some have
>>>> definitions *a version of which is being done by Chris Harding
>>>>   Some have categories attached to them.
>>>> I am not ready to share them publicly,  but I have managed to publish
>>>> one set *KRL in the proceedings of the DCMI conference
>>>> *should be open access soon  together with the evaluation methodology
>>>>
>>>> I have shared the vocabulary drafts on this list with CG participants
>>>> who expressed an interest
>>>> At this stage, and not until we have a better understanding of how to
>>>> handle this material, I have restricted access
>>>>
>>>> It s hundreds of hours of very interesting work
>>>> In the link below, shared on a post to this list on 22 October here, I
>>>> provide the rationale and pointer to the KRL vocab *corresponding to the
>>>> bubble on the right in the diagram
>>>>
>>>> *pre recorded talk here
>>>>
>>>> https://drive.google.com/file/d/1ZXGxsFXvnTXm_knYXbax0I2HdqfWuNWK/view?usp=sharing
>>>>
>>>> I do not expect that participants read all emails but that s the only
>>>> way to keep up sometimes
>>>>
>>>> This is why I plan a report shortly summarizing all the resources
>>>> produced to date, and of course, welcome evaluation and refinement
>>>>
>>>> PDM
>>>>
>>>>
>>>>
>>>> On Wed, Oct 22, 2025 at 4:48 PM Paola Di Maio <paola.dimaio@gmail.com>
>>>> wrote:
>>>>
>>>>> Gday AI KR CG
>>>>>
>>>>> Today DCMI 2025 Conference takes place in Barcelona, our paper was
>>>>> accepted for presentation *in person only
>>>>>
>>>>> *pre recorded talk here
>>>>>
>>>>> https://drive.google.com/file/d/1ZXGxsFXvnTXm_knYXbax0I2HdqfWuNWK/view?usp=sharing
>>>>>
>>>> KRL vocab
>>>>
>>>>
>>>> Some related work done in the last few years in this space is
>>>> summarized in a recent narrative
>>>>
>>>> https://drive.google.com/file/d/1uZAw90qx1tPDbnVfGBSRCK260q-i8tzI/view?usp=sharing
>>>>
>>>> Here the slides where you may be able to open the links from the
>>>> recording above
>>>>
>>>> https://drive.google.com/file/d/1uZAw90qx1tPDbnVfGBSRCK260q-i8tzI/view?usp=sharing
>>>>
>>>> *@TPAC 2025?  *https://www.w3.org/community/aikr
>>>>> /wiki/AI_KR_CG_@TPAC_2025
>>>>>
>>>>
>>>>
>>>>> A mailing list, you can browse the archive here, where we try to keep
>>>>> track of current topics in AI KR
>>>>> https://lists.w3.org/Archives/Public/public-aikr/
>>>>>
>>>>> Some working notes on the wiki * may be outdated!
>>>>> https://www.w3.org/community/aikr/wiki/Main_Page
>>>>>
>>>>> A first report published some time ago,
>>>>> https://www.w3.org/community/aikr
>>>>> /wiki/File:AI_KR_FIRST_REPORT_PUBLISHED_VERSION.pdf
>>>>>
>>>>
>>>>
>>>> Be good
>>>>
>>>>>