[uf-discuss] hResume skills property issues

Fri Sep 25 06:03:54 PDT 2009

Hi all

I would like to try and move the hResume microformat forward and deal
with some of the outstanding issues. I have placed the contents of this
email into a wiki page. Please feedback your thoughts on my suggestions.

http://microformats.org/wiki/hresume-skill-brainstorm

Tagspace issue.
When you review how skills are authored in resume they are equally
expressed either as terms or narrative sentences.  The following
examples where taken from a small sample [1] of online resumes:

* Object development
* CSS
* C# developer with 5 years team commercial experience 
* Quick to recognize problems and execute solutions
* English - Very good
* Speak, read and write English fluently

As Ciaran McNulty mentioned in the hResume issues page," in reality
authors end up linking to places like Wikipedia". You are also unlikely
to create a tagspace just to support the publishing of your resume.  I
think this is true whether you're adding a resume to a blog or building
a large resume hosting service.

On top of which tagspaces do not work well with the narrative style of
expressing skills used by some authors. Although rel-tag can create
interesting cross linkages it is no use if it cannot be practically used
for the majority of use cases when authoring a human readable document.

I would suggest that a simple text property is used instead of the
rel-tag.  This would be a more practical solution, as below:

   <p class="skill">Application development</p>
   <p class="skill">C# developer with 5 years team commercial experience
</p>

Separating language from skills
If you have read many resumes you will appreciate how varied the
labelling of skills sections are. There is one common division that I
think should be reflected in the schema, the separation of languages
from other skills.

>From a sample of 10,000 structured resume's entered into one of our
systems 42% entered a skill, 35.2% entered a language skill and 18.9%
entered a professional skill. This system is based in Europe and has an
international reach.  Please remember that structured entry systems can
encourage a bias that would not be seen in a sample of freeform resumes.
Even the small sample [1] of 10 freeform resumes picked at random had 3
that listed language skills.

We could redesign 'skills' into two separate properties;  'skills' and'
language' and use a common compound structure "competency"  to describe
both or we could use the type pattern to define the type of skill.

Using  a competency structure

   <p class=" competency skill">Application development > </p>
   <p class=" competency language">French</p>

So the hResume spec would have two properties; ' skills' and 'language'
which would be of the type competency, just like the contact property in
hResume is of the type hCard

Using a type pattern
   <p class="skill">
        <span class="summary">Application development</span> 
        <span class="type"> skill</span>
   </p>

   <p class="skill">
        <span class="summary">French</span> - 
        <span class="type"> language</span>
   </p>

I think that the type pattern would be very verbose with large lists of
skills therefore I would suggest the creation of the competency
structure.

Rating skills
Rating skills are very common, it helps to give the reader an
understanding of the level of competency claimed by the author. The
Microformats community already has a nice rating pattern in hReview that
can be applied to a competency structure. It allows a value of (0.0 -
5.0) to be applied to an item.

   <p class=" competency skill">
        <span class="summary">Application development</span> - 
        <abbr class="rating" title="1">Basic</span>
   </p>

   <p class=" competency language">
        <span class="summary">French</span> - 
        <abbr class="rating" title="5"> Fluent </span>
   </p>

I wish I could backup the use of ratings with a good set of stats. Our
structured resumes place an optional rating next to each skill entered.
The UI of these systems encourages users to rate their skills and gives
a bias to any results.  Consequently, 97.7% of skills in our sample
10,000 structured resumes are rated by authors. I did look at using a
Natural Language Parser to exact some stats from a collection of
freeform resumes, but the error rate on parsing skills ratings is high
and would not give a true result.  Therefore the only reliable stats I
have are that 2 out of my 10 randomly picked online resumes used
ratings.

Durations
It is common to express skills in terms of years/months of experience.
This can be industry specific and is most often found in IT. It would be
relatively simple to add an ISO Duration to the competency structure.
We already use this data format in hCalendar. 

   <p class=" competency skill">
        <span class="summary">Application development</span> - 
        <abbr class="duration" title="P3Y">3 Years </span>
   </p>

Again, I am unable to back this up with a good set of stats.  Duration
is less commonly used than rating, but is still important in certain
industry sectors. 

I would like to purpose a new competency structure to replace the
current 'skills' property in hResume.  We add two separate properties
skills and language that use a common structure "competency"

* competency 
  * Summary (text)
  * Rating (0.0 to 5.0)
  * Duration (ISO Duration)

   <p class=" competency skill">
        <span class="summary">Application development</span> - 
        <abbr class="rating" title="3"> Intimidate </span>
   </p>
   <p class=" competency skill">
        <span class="summary">Application development</span> - 
        <abbr class="duration" title="P3Y">3 Years </span>
   </p>
   <p class=" competency language">
        <span class="summary">French</span> - 
        <abbr class="rating" title="5"> Fluent </span>
   </p>

Implied competency optimization 
It would be worth considering a new implied optimization rule for
competency similar to the one used on the org property in hCard. Where
an author only wants to use a value without rating or duration the
summary property can be omitted and the value placed directly in the
parent element, i.e. both structures below would be valid.

   <p class=" competency language">
        <span class="summary">French</span> - 
   </p>

   <p class=" competency language">French</p>

[1] 
Madgex, the company I work for, hold approximately 4 million resumes
across a number of job boards we host for our clients. I can, under some
circumstances, do analysis of this data as long as it falls within the
use prescribed by EU/UK data protection laws. I cannot publish examples
from this data.  Therefore, I picked 10 random online resumes as a point
of reference. 

These resume have all been written freeform, come from different
industries and represent individuals from different nationalities.

Janne Kotiaho	 - Professor
http://users.jyu.fi/~jkotiaho/cv.html

Donald Henslay- Aviation Consultant
http://www.safeskyllc.com/don's_resume.htm

Jeremy Nevill	Software - Architect
http://www.nevill.co.uk/

Tristan Nitot - Company President
http://www.nitot.com/cv/

Lezli Renee (Thomas) Gill - Teacher
http://www.valdosta.edu/~lrgill/Resume.html

Lucie Moussu	- Assistant Professor
http://www.moussu.net/TP/curriculumvitae.html

Susan Sparkes-Hoskin - Business Development
http://www.telusplanet.net/public/shoskins/Sue_Hoskins_Resume_2008.rtf

Jeff Colen - Tour Promotions
http://www.jcolen.com/Resume.htm

Michael Barnett - Sales Training
http://www.michaelcbarnett.com/resume

Okwaisie Timothy - Metallurgist
http://www.min-eng.com/cvs/okwaisie.doc

Thanks Glenn Jones