Loading ...
Sorry, an error occurred while loading the content.

RE: [K-Logs] Digest Number 25

Expand Messages
  • Amy Wohl
    Metadata doesn t work very well because few people fill much of it out. And if we have it generated automatically, it often isn t very interesting or accurate.
    Message 1 of 1 , Nov 19, 2001
    • 0 Attachment
      Metadata doesn't work very well because few people fill much of it out.
      And if we have it generated automatically, it often isn't very
      interesting or accurate. As for the comments about Monster.com and its
      move to ask for more information I think it would be worth pointing out
      that we still have only about 2% of all job matches being performed via
      the Internet. So even when the stakes are relatively high (finding a
      job) people are necessarily very willing (or very talented) at filling
      in Metadata.

      We're starting to see a lot of AI-based search engines or other search
      technologies surfacing. Some of it is natural language processing
      based; others are more based on probability theory. In any case, I
      suspect that metadata isn't the only way to go and we're always going to
      do better when we can decide after the fact what we'd like to be looking
      for.

      Amy Wohl

      Editor
      Amy D. Wohl's Opinions
      Wohl Associates
      915 Montgomery Avenue
      Narberth, PA 19072
      (610) 667-4842
      amy@...
      www.wohl.com

      subscribe to our weekly Opinions newsletter FREE by clicking here
      www.wohl.com/signup.htm





      -----Original Message-----
      From: klogs@yahoogroups.com [mailto:klogs@yahoogroups.com]
      Sent: Sunday, November 18, 2001 11:39 AM
      To: klogs@yahoogroups.com
      Subject: [K-Logs] Digest Number 25


      To unsubscribe from this group, send an email to:
      klogs-unsubscribe@yahoogroups.com


      ------------------------------------------------------------------------

      There are 6 messages in this issue.

      Topics in this digest:

      1. RE: Will K-Logs be used?
      From: "Doug Kaye" <doug@...>
      2. RE: Will K-Logs be used?
      From: "Phil Wolff" <Philip.Wolff@...>
      3. Re: Will K-Logs be used?
      From: "Dann Sheridan" <arkangel9@...>
      4. Re: Will K-Logs be used?
      From: "John Robb" <jrobb@...>
      5. Re[2]: Will K-Logs be used?
      From: jotajota <jotajota@...>
      6. Re: Will K-Logs be used?
      From: M�r �rlygsson <mar@...>


      ________________________________________________________________________
      ________________________________________________________________________

      Message: 1
      Date: Sat, 17 Nov 2001 05:19:21 -0800
      From: "Doug Kaye" <doug@...>
      Subject: RE: Will K-Logs be used?

      jotajota wrote:

      "we cannot rely on the computers to find the relevant metadata"

      Do you really believe that search tools are inadequate? Have you
      investigated high-end tools such as neural-net/AI text-analysis engines?

      It's been my experience that (a) every metadata field one adds to a
      system such as this causes a measurable decrease in usage and
      acceptance, and (b) taxonomies created in support of such metadata
      always (yes, that's a strong
      word) collapse. Of course, a change in taxonomy usually invalidates any
      metadata entered previously.

      Obviously, I'm at a far end of the metadata spectrum. Particularly since
      K-Logs encourage the entry of true microcontent (thoughts, whims,
      sentences, links), ease-of-use is critical.

      ...doug

      Doug Kaye
      doug@...
      www.rds.com





      ________________________________________________________________________
      ________________________________________________________________________

      Message: 2
      Date: Sat, 17 Nov 2001 08:43:57 -0800
      From: "Phil Wolff" <Philip.Wolff@...>
      Subject: RE: Will K-Logs be used?

      Metadata in proportion to user motivation is a huge improvement. I'm not
      convinced that klogging with metadata kills the process.

      Career sites have had similar problems. �

      Well-structured descriptions of people and jobs slash noise, improve
      relevance of search results, and enable all kinds of value-added
      services. They have traditionally required effort, but that may not
      always be the case.

      In its early days as the Monster Board, TMP collected something like 6
      fields of information about you: 5 contact fields and a big blank box to
      paste your resume. This made it easy to reuse your resume and got lots
      of people to sign up; full text search was used to find candidates. The
      result was noise and inefficiency; fewer than 2% of those seeking work
      on Monster found it there.

      There have been several responses from the ecruiting industry.

      First, they increasingly ask you to pick descriptive information from
      lists. Health care job sites ask for your nursing specialty and
      subspecialty. Pick metro areas where you want to work. This quickly
      segments very large oceans of candidates/jobs (more than 10% of all
      Americans have a resume online) into more manageable pools.

      Second, they try to standardize frequently used descriptive language.
      Microsoft Word, MS Word, Word, Word 10 all refer to the same KSA
      (knowledge, skill, ability) area. But search engines usually fail to
      recognize this kind of synonymy.

      Third, they try to extract structured data from unstructured text. Save
      user time and effort if your software accurately puts text in the right
      field, for user approval. Some of these systems are incredibly smart in
      their domain. See http://dijest.editthispage.com/matching/dataprep for
      vendor lists and a little more on this topic. These make the case that
      it is easier to edit/approve than to compose.

      Fourth, domain specialization is rising. Universal taxonomies frequently
      miss the point. While half of ecruiting activity occurs on the top 5
      generic job sites (Monster, HotJobs, et al), niche sites are popular.
      They respond to particular occupations, industries, and metro areas.
      Surgical nurses of Seattle describe what they know, what they�ve
      accomplished, what they want, differently than surgical nurses in
      London, than pediatric nurses, than corrosion engineers. �It makes all
      the difference in the world when a site understands your business, your
      job.

      In 2000, Monster moved to moderately structured work histories and KSAs.
      Their traffic suffered a learning dip, but resumed quickly as motivated
      job seekers and employers adapted to the new system.

      When Manila added site structures and departments, usage went up, not
      down. I've had folks link directly to my site by department
      (http://dijest.editthispage.com/newsItems/viewDepartment$staffing) and
      by hand-crafted url (http://dijest.editthispage.com/tools/pm). With rare
      exceptions I almost always use the department functionality for
      day-to-day blogging.

      Metadata in proportion to user motivation is a huge improvement. I'm not
      convinced that klogging with metadata kills the process.

      - phil

      Philip Wolff
      evanwolf group
      http://dijest.com
      pwolff@...
      510-444-8234





      ________________________________________________________________________
      ________________________________________________________________________

      Message: 3
      Date: Sat, 17 Nov 2001 15:43:03 -0500
      From: "Dann Sheridan" <arkangel9@...>
      Subject: Re: Will K-Logs be used?

      I think Doug, Jotajota, and Phil have highlighted the key points to the
      success or failure of k-logging:

      1. Ease of use vs. data structure
      2. Synthesis of large amounts of content/data into a consumable format

      K-logs on a small to medium scale will be fairly easy to keep useful and
      relevant. By small to medium I mean groups of people in the 10's and
      low 100's that are focused on a common problem or are part of a common,
      immediate organization. Once you have more than a few hundred people
      that are part of different organizations or content/data begins bridging
      organizational boundaries, things rapidly get out of control. One big
      five consulting firm that I am familiar with spends millions of dollars
      a year categorizing, maintaining, and distributing their "knowledge
      capital" to teams in the field. The down side to this approach is that
      the field teams enter their knowledge _after_ the work is done, if it
      gets entered at all. There is no tracking of individual thoughts or
      perspectives, only the team's. To complicate things further, the
      content/data comes in every format imaginable: PowerPoint
      presentations, Word documents, Visio diagrams, text, spreadsheets,
      Lotus Notes databases, SQL databases, etc. There is always some meta
      data collected, but usually not enough to understand the contents of the
      contribution. At the end of it all the size of the problem ends up
      being some 3,500 relational databases distributed on six continents that
      need to be replicated at least once a day to hundreds of local offices
      and accessible by people in the field that may or may not be on the
      corporate network. If each person was able to contribute their own
      perspective and everyone had access to individual perspectives, the
      problem of simply distributing the data would be unmanageable not to
      mention structuring it or synthesizing it into a consumable format.
      They still don't have a good approach to making the knowledge accessible
      other than through a search engine that spits out results based on key
      words and phrases.

      On the other end of the spectrum, there are people who aren't part of an
      organization, like this group and many of us as individuals, that are
      constantly receiving content/data across organizational boundaries. Our
      organizational boundaries happen to be our own brains. For example, I
      installed Userland's Radio on my desktop at home the other night and let
      it run all day yesterday. I subscribed to six different feeds. By 7pm
      there was so much content/data that I couldn't wade through it all. As
      I am learning this new way of aggregating knowledge I will have to find
      a way to optimize the time I spend reviewing content that may or may not
      be relevant to my current activities or interests. My initial reaction
      was to unsubscribe from some of the feeds, which I did. Now, I am
      trying to find a way to categorize the content as it comes in versus
      having to manually wade through it once or twice a day. Despite the
      power of technologies like RSS and applications like Radio, I still
      don't have a data structure that suites me personally -- yet. I am
      working to remedy this. :) The data comes in many formats, despite the
      structure RSS establishes, rendering it unconsumable until I sit down
      and read it like a newspaper.

      I would like to issue a challenge to this group: come up with three or
      four dictums for each of the two points above that will meet the
      following objectives:

      1. Clearly establish a balance between ease of use and data structure
      under some common scenarios we all live through in our professional,
      daily lives. 2. Clearly define approaches to the synthesis of large
      amounts of data that can be easily consumed by individuals.

      To kick things off, I will assert that the flow of data within in an
      organization needs to mirror organizational structure in order to be
      relevant. I will further assert that there are three tasks of today's
      knowledge worker:

      1. Communication
      2. Knowledge generation
      3. Knowledge acquisition

      Based on these assertions, key points, and objectives, I will begin
      authoring more posts that beginning driving towards an end. I have no
      idea what that end is, but I believe we'll know it when we get there!

      I look forward to a very spirited discussion!

      Cheers,
      Dann Sheridan



      ________________________________________________________________________
      ________________________________________________________________________

      Message: 4
      Date: Sat, 17 Nov 2001 17:46:14 -0500
      From: "John Robb" <jrobb@...>
      Subject: Re: Will K-Logs be used?

      Dear K-Loggers,

      This is a great discussion. Here are my thoughts on Dann's assertions,
      given his Big Co perspective:

      Dann: The down side to this approach is that the field teams enter
      their knowledge _after_ the work is done, if it gets entered at all.
      There is no tracking of individual thoughts or perspectives, only the
      team's. To complicate things further, the content/data comes in every
      format imaginable: PowerPoint presentations, Word documents, Visio
      diagrams, text, spreadsheets, Lotus Notes databases, SQL databases, etc.
      There is always some meta data collected, but usually not enough to
      understand the contents of the contribution.

      Answer: K-Logs address this by providing people with a simple way to
      enter knowledge on a daily or hourly basis -- as it is created. As a
      result, little knowledge is lost. Also, given the "story-line" nature
      of a K-Log provides it provides contextual details necessary (required!)
      for understanding. For example: "Here is my PowerPoint proposal (link
      to PowerPoint) for the structure of client x's network. It includes:
      x,y,and z." and "The response from the client to proposal was: a,b, and
      c. As a result of this feedback, here is my revised proposal (link to
      PowerPoint)"

      The problem with most existing tools is that they are too structured to
      be easy to use. A general tool like a K-Log can provide immediate
      benefit without complicated interaction.

      Unlike any other solution available today, K-Logs can reach
      Internet-scale. Hundreds of thousands, if not millions of K-Logs posted
      to a common Intranet would be 1) inexpensive to build (relative to other
      solutions), and 2) a huge advance over the current situation, and 3)
      easy to both contribute to and get value out of. The potential that
      there would be loss in the transmission of the knowledge contributed can
      be lived with. This negative argument is akin to saying that the Web
      should be shut down because sometimes you can't find exactly the
      information you need. From another perspective: the free market is a
      buggy and messy system, but its the best system we have found so far.

      Dann: If each person was able to contribute their own perspective and
      everyone had access to individual perspectives, the problem of simply
      distributing the data would be unmanageable not to mention structuring
      it or synthesizing it into a consumable format.

      Answser: K-Logs are automated Websites (they also may be simple static
      sites hosted on the Intranet). If an office or an employee working
      remotely has access to the Intranet via a VPN (virtual private network),
      then distribution of the knowledge is accomplished without resorting to
      expensive replication techniques. Further, the organization of the
      knowledge is simple, and effective: K-Logs are based on individuals.
      They are also time organized. Finding relevant knowledge is easy: 1)
      you can use cloud-based knowledge network tools(
      http://groups.yahoo.com/group/klogs/message/62 ) to locate relevant
      knowledge elements, 2) you can subscribe to a person's knowledge stream
      if you think their perspective is valuable, 3) the contextual
      information you get from the individual's K-Log will help you understand
      how to use the data they post. How much simpler can it be??

      Dann: My initial reaction was to unsubscribe from some of the feeds,
      which I did. Now, I am trying to find a way to categorize the content
      as it comes in versus having to manually wade through it once or twice a
      day.

      Answer: There are a couple of ways to remedy this (so all you get is
      the good stuff):

      1) To have people to publish a K-Log to utilize categorization to build
      audience or project specific K-Logs. That way as a consumer of
      knowledge I can tap into a very specific knowledge stream.

      2) To build tools that let me utilize key-words and AI to narrow my
      knowledge streams. This can be done through a generalized search
      service (Moreover) or through a narrowing of a specific K-Log's
      knowledge stream.

      Let's be clear: subscriptions should only be used when the knowledge
      stream you are interested in is highly qualified. You shouldn't use a
      subscription to aggregate everything (at least not yet). Longer term,
      it is possible that more sophisticated methods of automating the
      analysis of knowledge streams will be developed. The key to this
      discussion is solving the basic problems first before we jump to
      describing the functionality of the mature solution.

      Dann: To kick things off, I will assert that the flow of data within in
      an organization needs to mirror organizational structure in order to be
      relevant.

      Answer: That has already been recommended here:
      http://groups.yahoo.com/group/klogs/message/62 K-Logs listed within an
      organization framework will over time become a feature of K-Log
      knowledge network services in the cloud.

      One last word about metadata: it is almost worthless. Why? People
      don't enter it. Also, people utilize different taxonomies (my metadata
      isn't your metadata) which cause misunderstandings. This is a
      technological dead-end and a big reason why Tim Berners-Lee's Semantic
      Web won't become a reality (but I digress). A big improvement over this
      approach is to use categorization to build specific knowledge streams
      based on the needs of groups of users.

      Remember, the world is a messy and relatively undisciplined system of
      interactions. We can only hope to moderately improve those interactions
      by introducing new technology. Start simple and let the system evolve.

      Sincerely,

      John Robb
      http://jrobb.userland.com



      ________________________________________________________________________
      ________________________________________________________________________

      Message: 5
      Date: Sun, 18 Nov 2001 12:45:10 +0100
      From: jotajota <jotajota@...>
      Subject: Re[2]: Will K-Logs be used?



      > Do you really believe that search tools are inadequate? Have you
      > investigated high-end tools such as neural-net/AI text-analysis
      > engines?

      Well, no. Can you give me some links about this? TIA

      jotajota




      ________________________________________________________________________
      ________________________________________________________________________

      Message: 6
      Date: Sun, 18 Nov 2001 13:34:53 +0000
      From: M�r �rlygsson <mar@...>
      Subject: Re: Will K-Logs be used?

      On 2001-11-17, 13:46:50 (+0100), jotajota wrote:
      > This is *almost* true. As the number of users and the amount of
      > information grow in the Klog also increases the difficulty to find any

      > discrete piece of knowledge in need. This must be addressed by
      > metadata use. If you set up a nice metadata system, with plenty of
      > different categories and stuff, the information will become more
      > accessible BUT then your first premise:
      > > 1) The tool must be EASY to use.
      > will fail to be accomplished.

      Good UI design can help making meta-data easy to create.
      By limiting the options available (e.g. a mostly predefined list of
      keywords, etc.) and making the Meta-data entry optional the barrier of
      use virtually disappears.

      Also good UI design on the Weblog level can help the users feel the
      immediate benefit of entering meta-data. If your blog-tool allows users
      to browse their own weblog feeed (and
      others') filtered by keywords and other meta-data they should quickly
      come value meta-data and enter it more often.

      Then I also think that every company that truly values it's knowledge
      network has to assign someone to "oversee" the weblog cloud. This
      someone could be a part time worker with a good sense for IA/meta-data
      or a full time worker with library science education. Anyhow, this
      person would have the job of reading all the fees and adding (or
      editing) metadata as he/she sees fit. Tending to the knowledge field
      like a farmer.




      --
      M�r �rlygsson
      ------------------------------
      mailto:mar@...
      http://mar.anomy.net



      ________________________________________________________________________
      ________________________________________________________________________



      Your use of Yahoo! Groups is subject to
      http://docs.yahoo.com/info/terms/
    Your message has been successfully submitted and would be delivered to recipients shortly.