“RELEVANT
SEARCH”
From:
Hemen Parekh
To:
Rahul → Saurabh → Pranav
Date: 12
Apr 2006
(With
a pasted Times of India clipping titled “Google gets advanced search code,” TOI
11-04-06)
Page
1 / 6 – Learning from Google’s Algorithm
“What
can we learn from this?”
You
note that:
- A
jobseeker is also searching for information — specifically, job
advertisements.
- Therefore,
Google’s principle of relevance-based retrieval can be adapted to job
search and even to resume search.
You
pose a key design question:
“The
text will only appear if ______ ?”
and
answer it:
“Obviously,
if those texts (i.e. job ads) contain keywords relevant to the search query.”
Jobsearch
UI Prototype
You
illustrate the Job-Search interface mock-up, complete with keyword input
and structured filters:
Job
Search
Keywords:
[___________]
Tip:
Ideal job-ad should contain these keywords.
Min
Exp asked for: [__] years
Industry: [▼]
Function: [▼]
Desg.
Level: [▼]
Job-location:
[▼]
[SUBMIT]
Below
the sketch, you add an empathetic behavioral note:
The
jobseeker visualizes an ideal job ad, scans and memorizes its keywords, then
enters them before the memory evaporates —
“All
these, so that our s/w can ‘match’ these keywords in the texts of job-adverts.”
From:
Hemen Parekh
To:
Rahul → Saurabh → Pranav
Date: 12
Apr 2006
Page
3 / 6 — Simplifying the Jobseeker’s Life
“This
is too much to expect from a jobseeker!
We
must make his life SIMPLE.”
You
identify a timeless UX truth: reduce cognitive load.
You
note how even small data-entry burdens (typing keywords) discourage user
participation, and contrast this with:
“It
is very easy for us — but almost impossible to replicate or copy by
Monster/Naukri etc. (another USP for us).”
Then,
you propose a revised JobSearch interface (your second-generation
layout):
Jobsearch
Please
show me job-ads which match following:
---------------------------------------------
Function: [▼]
Industry: [▼]
Desg.
Level:[▼]
Job-location:[▼]
Min.
Exp. asked for: [__] yrs
---------------------------------------------
Keywords:
[__________]
(Delete
from this box, those keywords which are not 'relevant' from your viewpoint.
Feel
free to add any, when you discover 'relevant' but which we have missed out.)
[Submit]
Auto
fill-up “Keyword” box
The
moment a jobseeker selects any (one) “Function” from the Function drop list,
our system will pick up (some 20/30) keywords which our Function Profile Graph
uses (to draw the graph).
These
will be the top 20/30 keywords — in terms of their frequency of
occurrence, those having highest weightage — in the descending order of
weightage.
Besides
truly “amazing” the jobseeker with this magical appearance of keywords, we have
made his life simple!
No
more excruciating mental exercise for him to “conjure up” a set of keywords.
He
is already presented with a set of keywords which are truly relevant to
“Function.”
Given a set, it is easy for him to add more, but it is also easy for him
to see what words are missing — consequently by their absence!
And
whatever new/fresh words that he adds to the set are very valuable to us.
We
will store these in a separate database, called …
“Jobseeker
Suggested Keywords” — we will store these against each “Function.”
Then,
we can think of modifying our Job Search UI as follows:
Job
Search
Function
⬇️
Ind
Desi Level
Job
Loc
Misc
App
Keywords
(Suggested by us)
(Most
frequently used keywords suggested by previous jobseekers)
⬇️
Seekers
Suggest
(Skill
words / Knowledge words / Attitude words)
Delete
from this box
Submit
Page
6/6
Also,
whatever new/additional keywords that jobseekers suggest/add in the box, we
will keep adding up their frequency with which they are being suggested.
Then
modify our keyword profile as follows:
|
Old
Keyword |
Weight |
New
Keyword |
Weight |
|
1.
Marketing |
0.03 |
After
Sales Service |
0.025 |
|
2.
Sales |
0.02 |
||
|
… |
… |
The
moment the weightage of any new keyword exceeds the weightage of the
bottom-most old keyword, then that new keyword will push out/replace the old
one.
So
now our system has become self-learning and tuned to Social Consensus
/ Audience Pull!
We
can do same with “Resume Search” UI also.
ahul
– Sarabh – Pranav
Examine
the words used to describe this technology:
- Snapshot
- Subset
- Vast
Storehouse
- Optimized
- Content
Density
- Captured
& Compressed Info
- Sample
- Millions
of Answers
- Google
/ Yahoo
All
of the above apply to a specific “Search Query” (as in Google/Yahoo) and the
Search Results.
Our
Function Profile Graphs too are such a “Snapshot” or “Photograph.”
(Sketch
showing bell curve with “Function – Sales,” total population = 18293,
sub-population = 265, and percentiles marked 30–90.)
Data/info
about 18,293 executives is “squeezed/condensed” into a small graph!
Hence,
graph has a very high Content Density.
And
someday, one of our Resume Search methods will involve:
- Displaying
the graph, based on a HR manager’s “Search Parameters.”
- Enabling
the HR manager to place his cursor on the 70th percentile, then dragging
to 90th percentile (thereby highlighting that area in between) &
clicking.
- This
will result in a short display tabulation containing one-line
summaries of only those executives whose favorable scores lie between 70
& 90 (in descending order too!).
Even
before clicking, the highlighted section has told him he can expect to see
results for 265 executives meeting his criteria.
Any
area shaded or graphically enhanced will indicate the number of executives
covered in that range.
12/04/06
No comments:
Post a Comment