Hi Friends,

Even as I launch this today ( my 80th Birthday ), I realize that there is yet so much to say and do. There is just no time to look back, no time to wonder,"Will anyone read these pages?"

With regards,
Hemen Parekh
27 June 2013

Now as I approach my 90th birthday ( 27 June 2023 ) , I invite you to visit my Digital Avatar ( www.hemenparekh.ai ) – and continue chatting with me , even when I am no more here physically

Saturday 30 January 2010

JOB SEARCH- R.I.P.


Shalaka, Alex

Displaying "Matching" job - advts the moment a candidate completes each STAGE of Registration (-so as to encourage him to APPPLY ONLINE - and, thereby "Market" our website amongst corporate)

Sending - out "Job Alert E - mails" to him every day, when "matching" job - advts get downloaded (thru RSS feed).

Re - arranging "Job Search Criteria" (-when he arrives on Job search Page after login) - based on
      . What criteria he had "used" in past
      . What advt. he had "viewed" in past
      . What advt. he had "Applied" in past

(A "Recommendations System" based on Click - stream analysis of his past behavior)

Someday (soon) JAM /"Apply Online" from mobile Implementing the above - mentioned features is a pre - requisite for generating a "Buzz" amongst jobseekers and getting traffic.

This would require capturing every click that he makes on our website. Only then can we compile his "database of intentions" This is better than providing a structured FORM called "MY JOB PREFERENCES".

Because "PREFERENCES" keep changing over time. These are dynamic - not static. As a person gets older/ rises in his career, his needs change. We must track these, interpret these (eg " (1) overall frequency of usage of a particular criteria  (2) Last one month's frequency of usage, etc).

We should give "Weight ages" as follows.

    . Overall frequency - 0.5 (50% weight age)
    . Last month's frequency - 0.5 (50 % Weight age)

For each candidate, we have to create 3 "click stream Analysis" tables, as follows.

Table NO. 1   ("Search Criteria" used Table)

Here we capture & store the frequency of usage/ selection of "Job Search Criteria" As shown in column 1 of Annex : A

Table No. 2 (What job Advts. did he open/ view?)

This is shown in column "2 of ANNEX A.

Obviously, he is not going to "open/ view" each & every job - advt that gets listed in search Table/ Search Results. Based on the data that gets displayed in search - results, he gets some more information about each listed job-advts. This table tells him something more about each advt.

So, out of these, a few advts. will "interest" him (i.e. those which SEEM to match his NEEDS"). He will click open these few to take a closer look to confirm that these DO really MATCH his needs.

So, at this (clicking) level, we get (capture) a deeper insight into his interests.

Eg. 
    . Which "Companies" Advt. he click - opens
    . Much  more frequently?
    . Which "cities" top in his search ?
    . What "Job Titles" attract his attention ?  etc. etc.

Table No. 3 (Finally, which Job Advt. did he apply ?"

Although he may open 100 job advts. (over a period of time), he may actually apply against only 6 of these What made him choose/ select only these 6, out of 100?

Obviously, when he "opened/ viewed" those 100, he found some "Additional Info" which were missing from the search result tabulation.

This "Additional Info" is shown in column of ANNEX : A.

What he "found" in
    . Job Description
    . Desired Profile
    . Keywords

rang a bell !

"Ah, here is my dream job !"

His mental computer quickly "indexed" all the keywords found in these 3 fields and (most likely) "Matched" these keywords, with the keywords he had listed in his own resume.
If he found a good "Match" between the TOW SETS of keywords, then that was his DREAM JOB.

Match Index = Raw Scored Of Resume / Raw Scores Of Job Advt.

The higher the index, the better the Match. But we know that

Candidates do not UPDATE their resumes at regular interval, to incorporate keywords pertaining to their RECENTLY ACQUIRED SKILLS. Resumes are obsolete !

Recruiters are LAZY / INCOMPENTENT and Just do not bother to incorporate the relevant keywords in the fields for
    . Job Description
    . Desired Profile
    . Keywords

Hence, to begin with, we will not bother to compute the MATCH - INDEX.

We will simple INDEX the words (all the words) Contained in each & every Job . advts. that this candidate "Applies" And add - up. And, rearrange these WORDS (not keywords i) in the DESCENDING Order’s, we will know, which WORDS are MOST FREQUENTLY found in all the jobadvts against which this candidate applied.

There is some "Co-relation" between these most frequently used words and the candidate's interest in those jobs (advts) which contain these words.

So, when we download (RSS feed) any Job - advt, we INDEX its WORDS. Here there is no "frequency" - Since most words are likely to have been used ONLY ONCE in any given (one) advt_although there can be exceptions. Then we compare as follows. If we find that any downloaded job. Advt. contains ANY of these 3 words, then we can assume that it may be of "Interest" to this candidate !

At this stage, I suggest that we do not worry about HOW exactly we will use these 3 Tables to recommended to a candidate some "Matching / Interesting" jobs.

At this stage, let us simply focus on

    . Capturing a candidate's CLICKS

    . Storing these CLICKS into 3 Tables

    . Arranging Cumulative totals of CLICKS into descending order.

These 3 Table are the FOUNDATIONS of our RECOMMENDATION System.

Once we lay a strong foundation (by capturing & storing the clicks into 3 tables) then we will figure - out the next step.

Hemen Parekh

No comments:

Post a Comment