CAPTURING JOBSEEKERS KNOWLEDGE

Rahul

11-04-06

Capturing Jobseekers’ Knowledge

(– to make AumMitra a self-learning software)

---

We discussed this today.

In IndiaRecruiter, a jobseeker has to identify 3 industries & 3 functions, where he claims to have strong background.

But,

Which of these 3 (industries & functions) are,

▶ most “relevant”? (Where he feels superbly confident to succeed)

▶ quite “relevant”? (Where he still feels quite comfortable)

▶ somewhat “relevant”? (Where he can “get into the groove” with some brushing-up)

---

Our existing “Submit Resume” form does not bring out these subtle differences/nuances between those 3 industries / 3 functions.

But we have a STRONG need to capture these.

To do this, let us modify “Submit Resume” form as follows:

What is your background in terms of:

| Industry | Function | Keywords |

[Dropdown fields drawn under each]

It is quite unlikely that you feel equally comfortable with your choices of Industries & Functions. Which do you consider,

▶ Most relevant

[Where you feel superbly confident to succeed]

Rank: [ 1 ]

▶ Quite relevant

[Where you still feel quite comfortable]

Rank: [ 2 ]

▶ Somewhat relevant

[Where you can "get back into the groove" with some brushing-up]

Rank: [ 3 ]

In the box below, please “Rank” your choices

(to help us recommend to you the ideal jobs)

My Ranking is as follows

Industry	Type Rank	Function	Type Rank
[ Dropdown Option ]	3	[ Dropdown ]	2
[ Dropdown Option ]	1	[ Dropdown ]	3
[ Dropdown Option ]	2	[ Dropdown ]	1

Cut & Paste your text resume in the box below

[Text box drawn here]

Once we capture the ranking, we will store these in our database, against the name of the concerned candidate.

That will enable us to create "sub-populations" of candidates:

▶ Industry-wise

▶ Function-wise

Next Step

For all candidates belonging to industry ABC,

add up all the keywords contained in their

"Knowledge Profile" boxes.

Then calculate

▶ “Frequency of Occurrence” of each of those keywords (probability of occurrence)

Since, each candidate has identified himself as belonging to

▸ Industry = ABC

▸ Function = XYZ

and, he has himself used/selected certain “keywords” in his resume (i.e., in Knowledge Profile box),

we can safely “assume” that these keywords belong to those Industry / Functions!

---

So, now, instead of one or two “Experts” deciding

“Which keywords signify/denote which Industry?”

“ ” “ ” “Function?”

… we get thousands of real experts (i.e., the candidates themselves) to certify this relationship

(between “keywords” on one hand, and

“Ind/Func” on other)

---

This is exactly the future path of Yahoo’s search engine, viz:

evolve a “Social Consensus” thru a large no. of USERS voting/ranking/rating on items’ importance/relevance to the “Search Query”.

(Like AUDIENCE POLL in KBC!)

More & more search-engines are adopting this technique to arrange/display search–

–results in the descending order of the “Rank/Score” awarded by previous visitors.

---

This method (of creating smaller sub-populations) will also dramatically reduce server’s/software’s burden of computing “Frequency of Usage.”

This is because total candidate population (of say, a million resumes) will now (possibly) get broken up into 30,000 resume sub-populations (30 of them!)

---

Within each sub-population’s “keywords”, quite likely, the top 50 (arranged in descending order of frequency-of-usage), will add up to 90% of the sum-total of probability (i.e., add up to 0.9 probability).

Subsequently, for plotting percentile-graphs, we need to use only these top 50 (or 60 or 40) keywords for matching/finding from next arriving resume, to give Raw score.

---

Fresh computing of “Frequency of Usage” taking ALL keywords in any given “sub-population” (of industry or function), may be done once-a-week (over the weekend?)

(Signed)

11/Apr/06