Ontologies and Taxonomies
Ontologies and Taxonomies
I search based on how I remember things.
Knowledge repositories need to have integrated knowledge reasoning engines. They need to be able to infer new relationships and knowledge.
Ability to create secure, easy-to-find places where I can store any and all types of information related to a particular personal, work or shared topic.
Ability to share things as easily as I share a conversation or invite someone to lunch.
Need the correct software to be there when I need it.
Need the software to automatically be running when it needs to be running.
Applications need to be easily composable with one another.
1. A reasoning or inference engine fully integrated into the WinFS stack ...ideally shipped with WinFS so that ISVs would have easy access to it. MS Access would be an ideal application for providing an end-user experience for creating applications that can reason of the rich knowledge networks that can be represented in WinFS.
See http://dotnetjunkies.com/WebLog/mwherman2000/archive/2003/11/11/3435.aspx.
2. Ability to search for PowerPoints with green backgrounds and pictures with shiny red metal and blonde hair.
See http://dotnetjunkies.com/WebLog/mwherman2000/archive/2005/02/26/56783.aspx.
3. An easy-to-use and configure shared space experience for plain old Windows users.
See http://dotnetjunkies.com/WebLog/mwherman2000/archive/2004/02/14/7232.aspx.
I've been thinking about all of the pictures I have and how I would really like to be able to search through them if they (image files) were stored in WinFS.
I'd like to be able to search the images to locate the ones that have blue sky or brown hair in them. How can I implement these capabilities in WinFS queries or MS Search (aka SharePoint Portal Server Search).
Is it something the SQL query processor would handle on the relational side?
...or is it something that MS Search could be extended to handle using a custom written iFilter. ...for example, a custom iFilter that can recognize blue sky or hair and knows how to save the appropriate properties in the MS Search “full text” index catalogs ...or with WinFS, will MS Search index catalogs become entities in the relational store ....all searchable in the usual way?
Another similar scenario being able to find all the PowerPpoint presentations with mostly black or mostly orange backgrounds.
Michael.
Of course, I had omitted the obvious differentation in my previous models: the ability to query/search objects and streams. :-) Here is my most complete rendition of the taxonomy of capabilities for WinFS (IMHO). There will likely be a sixth dimension ...perhaps something about the continoum of independent MS and ISV object models vs. a fully integrated and architected set of WinFS schemas. I have no idea how I'll draw the 6th dimension. “More news at 11...“
[Note: I have no official relationship with the MS WinFS team.]
Note the *K* in SRBK.
The previous SRB 2.1 version was achievable using any modern day Windows application that exposed a rich object model. How do I think WinFS will differ? It's the piece about relationships and WinFS's ability (as of the 2004 PDC) to represent arbitrary objects and arbitrary relationships and hence, arbitrary knowledge networks ...opening the door for new categories of every day applications that will use reasoning engines, etc.
Click here for this version of the diagram.
[Note: I have no official relationship with the MS WinFS team.]
Some day I'll have a specific preference for SRB or PSM ...but for now I'm floating both. In this version, I've replaced Accessibility with Storage ...as the concept of Accessibility overlapped significanty with the Structuredness/Properties axis. Storage is a better choice.
Click here for this version of the diagram.
People have asked for more detail about what the axis represent in my original diagram: http://dotnetjunkies.com/WebLog/mwherman2000/archive/2005/02/08/52608.aspx. See below an updated version of the SRB trinity diagram.
An additional key thought is that there is a multi-dimensional continium between the simple through rich data formats/objects that live outside the WinFS store and the simple through fully supported structured data objects that live inside WinFS.
Cheers,
Michael.
Click here for this version of the diagram.
In reading http://weblog.infoworld.com/udell/2004/07/20.html (dereferenced from Jon Udell's Office DevCon report @ http://weblog.infoworld.com/udell/2005/02/04.html#a1168), I was intrigued by the synergy evident in the ROX trinity (relation/object/XML) that Quentin Clark, director WinFS program management, used to describe WinFS's support for WinFS objects (structured data), XML (semi-structured data) and unstructured data.
But what about other existing, every-day data/stream formats such as vCard contact infomation, LDIF (LDAP Data Interchange Format) schema and data files, CAD (computer-aided design) or seismic data? These are characterized by being non-Office (non-OLE structured storage) and non-XML but they are definitely structured and usually, streams. Where do they fit into the ROX trinity (and WinFS)?
Besides WinFS objects, XML and unstructured data, I believe there is a need for a fourth category in WinFS: “stream structures” and/or “stream datatypes“. These are needed to embrace existing, every-day data formats like vCards, LDIF and seismic data, etc.
The difference between a stream format being a stream structure and stream datatype is the extent to which its implementation (including property access and behaviors) is supported in WinFS.
If WinFS only supports property access/promotion/demotion features for a particular stream format, then I would suggest using the term stream structure for the format. If the WinFS/Yukon CLR support is used to add WinFS support for rich collections of behaviors and methods for a particular stream format, then a stream structure graduates into being a stream datatype.
To follow a consistent naming strategy, I'd rename unstructured data to simple streams: streams that WinFS does not provide a) any native or ISV support for property access (the pre-requisite to graduating to being a stream structure) nor b) any native or ISV support for a set of methods/behaviors (the pre-requisite to graduating to being a stream datatype).
Where does this leave the ROX trinity? What bothers me is that ROX doesn't map back to the "objects (structured data), XML (semi-structured data) and unstructured data" model very well.
If you consider the
- structured data (objects)
- stream datatypes (including XML with behaviors)
- stream structures (including XML without behaviors)
- simple streams
model I'm proposing, the trinity is really about SRB: Structuredness, Representation and Behaviors (or alternatively, PAM: Properties, Access and Methods) ...and the extent to which a particular data or stream format is supported in WinFS (native MS or ISV provided).
These 3 dimensions of the SRB (or PAM) trinity is illustrated in the following diagram:
Click here for this version of the diagram.
The SRB Trinity fully embraces traditional data and stream formats (regardless of their level of support in WinFS) as well as providing an effective framework for characterizing the different categories of data formats that have full WinFS support (structured data and stream datatypes) as well as partial WinFS support (stream structures with only property support) or no support (simple streams stored a blobs).
Cheers,
Michael Herman
Parallelspace Corporation - Developers of collaborative business solutions for Microsoft SharePoint Corporation - Developers of collaborative business solutions for Microsoft Microsoft SharePoint Products and Technologies Corporation - Developers of collaborative business solutions for Microsoft Microsoft Microsoft SharePoint Products and Technologies Products and Technologies Corporation.
Just found this on Amazon ...I hope RDF (and XML) are still relevant in 2020 ;-). I left a review comment on Amazon (the first one) expressing the same sentiments.
WinFS's 5 Ses
- Schema
- Association
- Storage services
- Synchronization services
- Search services