June 28, 2005
personal ontology
Yesterday we had “Prime Tags”.
Today we have “Personal Ontology”.
From Forward Markets:
An ontology as defined by dictionary.com is:
“An explicit formal specification of how to represent the objects, concepts and other entities that are assumed to exist in some area of interest and the relationships that hold among them.”
A personal ontology by extension is simply one persons ontology as opposed to a global, all encompassing ontology (e.g. the kind of grand unifying and restricting taxonomy-like structure that Sig is railing against). A personal ontology on the other hand is more relevant to the individual.
I’m assuming this to mean: as with the blogosphere and links, if you give people tags, the organisation will build itself.
And if the organisation builds itself, then why do we need SAP telling us what to do?
Hamish? Sig? Feedback?
"Hugh's Daily Cartoon" Newsletter.
A new cartoon sent out every weekday morning to your inbox [RSS version here.].
A wee chuckle to start your day off right etc.
You may find something I wrote last year interesting.
http://www.nooranch.com/synaesmedia/wiki/wiki.cgi?WikiPoweredArgument
In your post you talk about personal ontology and then the organization building itself. If the latter happens then does personal ontology have to give way to an organizational ontology?
Let’s get practical. Does a Purchasing department have to standardize their tagging so that they can run through their daily procedures as fast as possible. What happens when the Accounts Payable department needs to look at purchase orders or vendors? Does AP slap their own ontology on top of Purchasing’s? How does AP know when to slap on some tags to a new purchase order or a new vendor?
I don’t think organizations will form themselves with personal ontologies, rather each person would have their own view of the organization. Unfortunately organizations are primarily designed as a structure for responsibility. Everyone needs to know who to blame when the shit hits the fan. With a personal ontology that would be pretty much impossible.
Now, this is not to say that the concept of a personal ontology for structuring the world is a bad thing, I just don’t think it is applicable to business structures.
My (incomplete) thought was that if I have my ontology and Sig has his and both are sufficiently robust in breadth and scope, then software can infer where and how they intersect.
Then, I proceed to contextualize what I publish with my ontology as does Sig or anyone else. When I search or when my handy dandy personal RSS agent searches on my behalf, my preferences are expressed in terms of my ontology and software takes care of figuring out if someone else
MarkN,
Such unbridled enthusiasm! You should be commended.
personal ontology
gapingvoid: personal ontology:
I’ve been working with Bayes-related stuff for a while now, and there actually is a lot of potential there for at least helping to figure out where different “tagworlds” can/should interface.
I was also talking to someone recently about ways to easily tag things you find online, and then share the tags locally (among friends, within a company) and/or publicly. (Not just content you create, but anything you find online.)
I think the trick is somehow to make it very, VERY easy to do the tagging. The tag sorting and arrangement can be difficult as long as it’s behind the scenes, but the UI at all levels has to be dead simple or people won’t do it.
I’ve figured out at least one way to do this, but it’s still way too “power-user-only-ish.” Need to think more about the non-geeks.
I’m not sure that the use of a personal ontology in a business context is impossible. In the given example, the Purchasing department may indeed have a ‘standard’, out-of-the-box ontology, but rather than being prescriptive it is just a starting point. Individuals can then layer their own ontology over the top of that, similar to the concept of “Favourites” in a menu system, which allows them to work a bit differently to the person in the next cube, but without expecting them to work it all out completely from scratch.
This would allow considerable flexibility in the organisation of work, and the processes used to accomplish. If all the ‘personal ontologies’ are shared, the organisation ‘builds itself’ as an aggregation of all the tagging that individuals do, and ultimately the individuals carve out a path that accomplishes the business objective WITHOUT setting in concrete the “one best way” of doing it. An efficient ‘path’ or ‘flow’ (to borrow Sig’s words) will become obvious from the most-used ontologies, but nothing prevents a change to that flow over time, and nothing forces an individual to conform to a single alternative (which would preclude improvements). Changes can be better managed, can happen concurrently without disrupting the existing flow, and can achieve better take-up as early adopters ‘prove’ that the change is beneficial (or the opposite — prove that it is not useful BEFORE it does too much damage).
As Anthony says I think the key is that every person has one VIEW of the information available in the organization.
How do you do that? Well you ‘Project’ a Context into the full information store and you get given a view for that Context.
A context is made up of both Implicit bits (User, Agent, Time etc) and Explicit bits (What is specifically requested, for example which tags are being filtered on or even which even URLs are being requested). Tags are very exciting but by themselves I don’t think they will replace the SAP’s of this world, the real challenge is to provide Context.
links for 2005-06-29
Future Tense: Exploring the Future of Work Group blog on the future of work (tags: future work blog corante enterprise) Grokster: Why You Should Care | Bayosphere Another sad decision by the supreme court (tags: law sharing) Google Earth…
Well, truth be told, the aim of SAP isn’t really intended to support a personal tagging or metadata approach, but rather to avoid it. If you have 10,000 people who cannot agree on what something is called, then you have a problem. SAP is not intended to be a translation mechanism for everyone’s individual world view, but is intended to impose a single world view on a large group of people. Now, you could argue that this is a suppression of individual creativity in some cases, and I might agree.
However, markets in general work because there are standardised contracts in the market, that are known quantities, fungible, repeatable, etc. This is important to the maintenance of many supply chain activities. It is also important to financial and commodities markets, shipping, and many other things. In this context the need (understandable) of the individual to have things in a personal context is irrelevant.
Ontologoy as personal approach is fine, but in a discipline like science or shared markets what you need is more of a Linnean classification that can be agreed on.
Now, if you are working to organise things that are internal to your effort, and you have the luxury of being the first in the market, and making the definitions, then this is fine, but the rub is when you have to conform to the market definitions.
Where I do see value is a meta-classification, which allows objects to be defined with the tags of different overlapping semantic sets. For example, if I am using one definition of risk for a financial instrument, and I am using a different view for the accounting, then it would be useful to have a set of the main tags for the two areas (pre)defined, that I could apply as a super set to the one shared object, to allow both risk and accounting views to be applied.
So, in short, what is the value of SAP? For one, it is the fact that we have existing captures of all of this metadata, and the market share to impose our definitions where no acceptable standard exists. (In addition, we write software that supports this.)
From the point of view of the flexibility, then we have the idea that the functional block and the data model are standard, but the sequence which you would like to run the process blocks together is non-standard, i.e. NetWeaver, or the reportng is cross functional, i.e. Business Warehouse. If you want to see more about these two things go to http://www.sap.com
If this means nothing to you, then likely you do not need SAP…
What you are saying Hamish is that SAP would have a problem, because IT can’t handle 10,000 potential views of reality. No beef with that, but SAP’s problem with the concept doesn’t invalidate it. In fact 10,000 individuals will probably gravitate towards semantic agreement, as there would be a clustering of related tags around common understandings.
You are still talking of ‘imposing definitions’ — that will be the death of monoliths like SAP and Oracle’s Fusion (and all its current predecessors). Who defines an ‘acceptable standard’ — you, Hamish? An acceptable standard is one towards which the 10,000 individuals gravitate, and it will be acceptable only until a more acceptable viewpoint grows from the ‘wisdom of the many’ — and towards which the 10,000 individuals will gravitate .…… you get the drift. The idea that an acceptable standard can be imposed, either by a software vendor or by management who live by the status quo, is rapidly approaching extinction. Now I’m sure the beast will live long enough to ensure your comfortable retirement, Hamish, but extinction is visible.
I don’t think the imposed ontology (eg, SAP) necessarily conflicts with the personal and/or “swarm” types.
If in my business there is a thing standardized as “WidgetFoo” then it’s important that I know it’s WidgetFoo and why. But that doesn’t in any way preclude my also tagging it as “BeefSteak” — if that helps me organize my information world, so much the better. And if ideas start to cluster around BeefSteak in an interesting way, then it’s time to examine how BeefSteak and WidgetFoo may be related.
Think about a city map. It’s very important that everyone agrees, more or less, on the official street names and numbers and neighborhoods and so on. Without that, you would have chaos.
But then also think about how real people think of that map. If you live in a city, you probably know exactly where dozens of places are without knowing their exact addresses — in some cases without knowing the street names.
“The ice-cream place in district 5, next to the wine bar where the students used to hang out.“
“The beer garden behind the movie theatre.“
“The green house about a block down from Szabi’s place.“
Taxi drivers are a good example of aggregators of this kind of info — if enough people know about “Joe’s, by the bridge” then the taxi driver will process which Joe’s, which bridge, etc. and “translated” it to the map. Someone wants to go to Joe’s, he asks, “Joe’s by the bridge?“
Overlapping personal ontologies — and groups of them — with “official” views of the world has been going on for millenia, it just hasn’t been implemented well in software (yet).
hmm… off to that ice-cream place now!
Combining a structured taxonomy with author-generated tags (personal ontology) — aggregating both, then comparing and contrasting the two sources (whether automated using algorithms or by well-designed browsing .. or both )
could conceivably provide a well-filtered source of information that welds together a corporate or domain-area focus with individual perspectives and analyses
I think
The point about having a standard definition is not that it is the only workable one for your personal context, but that it is often the only workable one for individuals who haven’t met or otherwised discussed the subject. If I call up and Audi dealership, and I say that I have broken the back rear light, then the dealer does not call up the main supplier, and say,
“Hey, can I have a rear light for Hamish’s car?” They say, “do you have a DF-1200 – 987? Can you send it to dealer A400?” Then there is a common, albiet externally imposed semantic, that is not just necessary to the conversation, but indispensible. Not just because the part number says twhat is required, but because the dealer number gives an address, a credit rating, a confirmation of payment, and other information that would otherwise be cumbersome and inexact in the transaction.
Now, if you ask me a question such as “would Google eventually give me a statistically based definition of a word or a translation?” (Something that they are working on at the moment,) then the idea of statistically based tag is very relevant. They are using it to determine statisically based translations, such as reading all available human translations of a work from English to French, and mapping simply on the basis of frequency, what a human would have used for that word, with some context presumably. This is a good way of generating probabilty based information.
Some IT systems are intentionally specific when it comes to definitions, and others not. My view would be that you can superimpose tag-like definitions on a structured world-view quite easily, but not vice-versa. (The Audi car part DF-1200 – 987 supports a tag set of “light” “rear” “driver side”, etc, but the tags do not themselves lead you to DF-1200 – 987 until you have so many of them that you have reduced the parts in the car to a single instance of each tag combination, and even then you are never sure. Because, basically, unless you have a detailed breakdown of the car parts from another source, you do not know if your tags describe the entire universe of the Audi car parts. Example if I forget the tag “passenger-side”, then I have no way of discovering or describing the car part for the other side of the car from first principles. In other words, tags are useful for analysing an existing data set but do not lead you very far in inferring other logical relationships.
Hamish, “DF-1200 – 987″ as model id is not very human readable, but at the end of the day it’s probably a product of something — “DF” being Audi A4, “1” a light, “2” back, “00” right or whatever. Tags as they are.
, any number of precise or iffy or surplus or detail tags can be added without loosing the original id.
It could just as well have been tags in human readable format. “Audi A4, light, rear, right” + supplier, year delivered, batch etc. (getting the precise object id now).
So far no real difference except for the ‘search’ string: “DF-****-***” vs “Audi A4” if you want all parts for that car model.
Here’s the difference:
The “classic number structure” uses sequence and would in all normal instances be limited to the fields (as in a SQL table).
The “tagged structure” is sequence independent (start with what comes to mind first — “rear”), it requires no “wildcards” — and most important — in a system built for that (see my ‘experiment’ the other day
Or, going radical and completely silly here, if you had no “tagging manual” the handler receiving the part (or the supplier) tagged the part freely and enthusiastically: With enough tags one should be able to find the precise part anyway — and without having to follow the logic of any “tagging manual” only your own fancy and logic!
Waddyasay?
Hi Sig!
Hmmm. Not sure. I agree that if you keep tagging long enough you can give a human friendly search mechanism, but I worry about a couple of things.
1. I come back to the “tag-not-imagined” problem. If I can describe lots of things that I can think of, I can reinforce and search my metal model of the car, and that is useful, but there still has to be discovery process to be sure that you have captured all the attributes. If I cannot imagine it, it does not mean that the car does not need it? That doesn’t rule out the useful of tags for search and retrieval, but I think that you would have to have a support mechanism for the original descriptive process.
2. Tagging is flexible, but it is iterative, and some complex problems are best tackled in a linearly analytical way. Only some, I hasten to add. The cut over point is when the iteration is going to take as long or longer to produce an evolved result as the linear analysis. This is less of a problem if the tagging will autogenerate, and this may be where a machine based approach is useful, like Googles language translation. An alternative is to use the tagging approach to map information that never HAD a structure to begin with. (Then it works very well.)
3. (Yeah, I know, I said two.) To my mind the role of tagging is to associate standard data formats that otherwise are not easily handled. For example, if you are trying to have both an Audi part number, and a Toyota part number in the same system, then the “light” “car” rear” “passenger” “Make” Model” “Year” is an imposed taxonomy on other data. Car parts aren’t that good an example, but many engineering parts are standard, but are then refererred to by proprietary system numbers. This would allow reuse.
I am still of the mind that in the case when there is a need for information exchange between two previously unrelated external parties, then the only way to manage this would be to impose standard tags, or use a previously agreed and imposed standard, like EDI or SWIFT, etc. I suppose arguably that is what in fact many of the XML business standards intend to achieve?
Not arguing agsinst the usefulness of tagging more that the tagging approach is not in itself sufficient for all circumstances.
Hi Hamish!
and thus we may misunderstand each other. Better then if she used not only her “tags” but also “tags” that I could relate to, and vice versa of course!
— I think you may use “iffy-blobs-intersecting” (tags superimposed) and give as good or better results and at the same time free people from manuals, training, endless discussions about standards, lawyers and deliver world peace 
Let’s fill up Hugh’s storage capacity, here’s another long one! (this is too good a discussion to stop because of Hugh’s blog-storage budget!
“Captured all the attributes” — not always necessary says I. “Tartan, skirt, male, Macleod, wear” should pinpoint Hugh’s Scottish attire (he says he has no such though), even if it far from captures all the attributes of such a fine garment!
“Descriptive process” is another crux — that is something quite different from person too person.
Try this: Describe a “Campagnolo Record crankset” by tags that comes to your cyclist’s mind, as many as possible. Then let your wife do the same (non-cyclist I presume?) and add those.
Her tag-set would not ruin your tag-set, but it would (especially in combination with yours) make it easier to “find” the object by somebody of her mindset.
Same issue with “linear analytical way” — my wife is definitely less linear than I (seems to me sometimes
(This makes a case for letting all add tags to any object… posts in a blog for example… note to self, will add that to the “experiment”.)
In other words, let the descriptive process be free from constraints and let the “object-seeker” free to apply his/her own logic without having to assimilate the logic of the “object-creator” or read the “manual” (as the employees of Audi has to do)!
“Exchange information between unrelated parties”, well, then you would have to exchange “manuals” (or impose and distribute standard tags) first, and that is what is lacking when people get on TV debate programs… and that’s why we have lawyers
Apply the “Campa crankset experiment” from above to this too: Ask your aunt if she understands precisely what you describe if you use only “standard cyclists lingo and terms”, then let your wife pipe in with her “tags” and see if not that would be helpful!
In the end it seems we disagree (for the time being
Of course, the current systems thinking and software would choke… but so what I say
The Yin and Yang of Tagging: Specification vs Interpretation
As I monitor the ongoing discussion of tags, prime tags, personal ontology it occurs to me that there are two sides to the practice and practicality of tagging. The first side is the specification of tags.
Hi Sig!
I actually agree that tags are fully useful. No doubt about it. My issue is that if I described the crank set, I might mention useful things like number of sprockets, any mounting pecularities, units of measure, and so on. My wife, who is not technically illiterate, but not a cyclist, would not. Now, for her purposes, her description is good enough, but for my purpose, it would be inadequate.
Now, from a subjective point of view both are valid, but from an objective point of view, a non-expert may be incapable of creating a view that would fulfill the needs of all users. Thus standardardised tags, I guess Hugh’s prime tags, or let’s use an un-PC word, expert tags, are needed… I was thinking about this last time we spoke, and I mentioned a tag directory or meta-format.
Ehh. Maybe I spent too long locked up with German Phds. =-)
Hamish,
Add your wife’s tags to yours, both sets are the tags for the crankset, and for good measure ask your Japanese neighbour (if you have one) to add a few more!
no, no
Then you have the beginning of a good description for the crankset that will be adequate for hardcore cyclists, the baker on the corner and perhaps a few Japanese as well! Not to say the id/name would say a lot more about the stuff than “REC-987262.PE.098″!
(That’s after all what Linnaeus and the Periodic Table try to attain, adding implisit knowledge!)
Then you may avoid having to exchange manuals, agree on “standards” have governments “standarization boards” and whatnot…
German Phds… hehe… Albert Einstein was one (until he became Swiss that is) and he certainly broke off from the standard thinking
Sig,
I agree with Hamish, here.
How do you add your wife’s tags to yours exactly?
Is she sitting with you when you tag it? Because if she isn’t I can’t see how she can add her tags, she won’t be able to find your “Campagnolo Record crankset” because until she tags it with her tags she can’t find it, and until she can find it she can’t tag it!
You certainly can’t tag it with her tags for her: you don’t know what her tags are, and if you think you do you are setting your self up as the ‘expert’.
This is a classic Chicken and Egg problem.
Hugh,
How do you feel about being a facilitator? Seems that your visibility to a broad spectrum of readers is bring people who otherwise wouldn’t talk together.
21 and 19 comments respectively is a little more than normal right? Perhaps because you have stumbled across a topic that needs a forum?
I imagine one could add their wife’s tags to their own simply by using something like del.ici.ous…
The thing is, all of this is pretty feasible stuff. You just build it and watch it work… or not work.
It’s like the blogosphere.… it’ll just happen (or not) by itself.… no big company’s pre-approval needed.
I think there is certainly big company action taking place in terms of outfits like Technorati or del.ici.ous trying to establish themselves as the defacto aggregator/mediator.
The nature of the beast right now is that such mediation is required both technically and financially. RSS is after all just a file full of data sitting on a web server waiting to be downloaded. It far from a proactive thing. Zealously tagging and metadataizing the shit out of it isn’t going to make it any more proactive.
For proactiveness nowadays we have to rely on advances in RSS reader sophistication or on ever more sophisticated aggregators like Technorati. I am not a fan of either of these alternatives, although the realist in me knows that this is how its will be.
This is why I appreciate the fact that folks like Sig are trying to push the envelope out to where I am.
Islands of data
Evolving connections abound
Meaning for all
Hugh,
I don’t see how something like del.icio.us helps you add the right tags for someone else.
You can add your tags to things you find. But how did you find them? To do that you need to know the tags which someone else used?
Saying something is pretty feasible doesn’t make it so.
Agreed tags are valuable and allow you to bypass a company’s pre-approval. But lets not get too carried away here: There is a big difference between finding the exact something you need (as in business) and finding something that might be useful (as in del.icio.us).
Tags are at their best when used as a way of re-finding ‘found things’, or of discovering things other people who ‘tag like you’ have found.
I am not saying tags aren’t useful, if I thought that I wouldn’t have spent 3 years of my life trying to create a tag based file system.
Let me make this real simple:
1) top-down hierarchies / categories are good for uniquely identifying and referring to nouns (physical objects and locations).
2) tags are good at creating and shaping the “meaning” of these nouns…Your personal *schema* for an object or idea if the tagging is local to a system you solely control (like a personal blog)…The social “conversation” if tags are contributed by the crowd and viewed in public.
3) when folks try to make boring inanimate objects “mean something” with tags – or just as bad – attempt to predict the future and impose their intepretations of reality on others via categories – bad things happen.
Rules of Thumb: a) if you can have an interesting conversation about it, tag it. b) if having a conversation about it might adversely effect your chances of getting laid, slap a category on it and get on with your life.
Just back to Hamish’s comments (I’m not realy picking on you Hamish!) about his Audi rear light. The search by tagging gives you a smaller list every time you add another tag, so if you neglected to tell the search that you were looking for a LEFT rear light for your Audi, then you would have a list longer than one item, but your item would be on it. You only have to specify enough tags to get to a workable list; a list short enough for a ‘sequential read’ where enough information is visible to uniquely identify your item — like all your wife’s tags for the tail light perhaps!
If anybody could add new tags, then ultimately the item would become more readily available via different routes. If the first person (after the object-creator) found it by ultimately doing a sequential read through a hundred similar items, and tagged it with things that would have helped him (or her in the wife’s case), then the next person will probably find it in a list of fifty — and add some of ‘their’ tags — etc, etc.
Having said that, BTW, I agree that for most things there would be an original set of tags that would match existing categories to a large extent, if only to make it easier for people to change or augment existing search methods and habits.
imprecision it is (tags continued)
Excellent, fun, make-me-smile discussion going on here! Foot-in-the-mouthishly I suggested that voluminous and imprecise tagging would do the trick for any kind of object or issue, leaving precision to a delivery method of iffy-tags-interception. That …
Tagging — search is the problem
There’s a lively debate going on between Hugh, Sig and others too numerous to mention on the tagging thing as an alternative to tree-structure style databases. The discussions are interesting in the context of new business models but that’s it.…