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

Friday, 28 October 2016

AIR ( part TWO )

AIR  ( Part Two )  =  Artificial Intelligence in Recruiting

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

A few notes that I made in margins of a book ( written in 2006 ) that I read in 2007

May be by now ( in Oct 2016 ) , someone has already implemented the type of  EXPERT SYSTEM  that I conceived in my notes

If not , here is a great opportunity for some Indian Start Up !

I would be happy to guide , if requested

hemen parekh

hcp@RecruitGuru.com

28  Oct 2016


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

Name of Book :  Artificial Intelligence Application Programming ( second edition ) /  2006

Author            :  M Tim Jones

When read      :   2007

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

Page   3

We are planning that " Weights " of each keyword gets automatically adjusted / updated dynamically , with arrival of each new resume in our database . We only provide the " seeds " - the starting weights


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

Page   4

Inputs are " keywords " / Outputs are " Raw Scores "


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

Page   7

" Staffing Requirement Prediction System "

This would be of interest to us .

Job Advts posted ( on any job site ) are nothing but expressions of a Company's " Staffing Requirements "

If we have 5,000 job advts for a given Company ( over last 3 years ), we should be able to find a time-series / trend and should be able to predict its " future " requirements

Simplest is extrapolation of past trends !


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

Page   46 - 47

We have talked of a concept like " Rubik's Cube " , to place ( ie bring together ) on each " Face " , keywords belonging to a given " Skill " or " Function "

Goal = Getting all " same coloured " keywords ( squares ) onto the same face of the Cube

Path = Time taken ( shortest )

Each " face " of  our virtual " Rubik Cube " would have SAME colour , if someone succeeds in virtually " Rotating " the layers along the 2 axis of freedom until all squares are SAME colour - meaning, that all the squares contain keywords belonging to SAME SKILL

We can " time " the game , to see which visitor ( who ) , manages to get all keywords ( squares ) on the SAME face in the shortest possible time - then give him recognition / credit by publishing his name on our web site

This could be a lot of fun - and could possibly draw a lot of young kids to our web site !

There will be no log-in / registration for playing this game . Just walk in and play .

Then download the COMPLETED / SUCCESSFUL image ( with time taken ) & email to all friends to prove your cleverness !


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

Page    69

Addition / deletion of " which " and " how many " keywords would result in a Raw Score better than " best found " solution / resume ?


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

Page    70

What keywords ( elements ) are always found " together " ?


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

Page   72

Is one resume a " solution space " in which particles ( ie Keywords ) are swarming around ?


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

Page    91

* Master set of 5,000 keywords

*  Function wise sets of 100 keywords

*  " Skills " and " Functions " are classifications


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

Page   92

*  Read my notes on " Expert System " / Eg : Finding keywords which have NEVER occurred before in ANY resume

   How will s/w KNOW that it is a NEW keyword ? and then KNOW , to which NEW skill / function does it belong to ?

*  Clusters are Functions / Skills / Cities / Designation Levels / Edu / Exp etc

*  Obviously WIPRO / INFOSYS / TCS ? Satyam , belong to a well-defined " Customer Set "

    They all have " common attributes " ie:

>  Similar / identical job advts posted

>  Similar / identical resumes searched


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

Page    93

*  Each " subset " can be , resumes belonging to :

>   Same  FUNCTION

>   Same   SKILL

>   Same  CITY

>   Same  EDU...etc

*  Corporate are " purchasing " resumes from our web site and we will have exhaustive data on their " purchases " , ie:

>   No and Types of resumes transferred to Folders / Opened & Viewed


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

Page    98

*  We too are planning a " Recommendation System " - which would recommend Job Advts to Jobseekers and resumes to Recruiters

*   If WIPRO HR manager short listed / viewed / interviewed , such and such candidates , same could also be of interest to INFOSYS  HR manager ?


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

Page   113

*  I have written some rules . 

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

Page    142  (  Chapter on " Ant Algorithm " )

*   Can we develop an " agent " ( ant ) for each Corporate Subscriber's , each " job advt " , which travels to " distant places " ( different job sites ) and safely bring back " Food " ( resumes )  to its own " Nest " ( folder ) ?

*   Millions of Ants ( s / w agents ) let loose on Web Network , each programmed to find a " specific " type of resume - then , when it finds , passing on this data to the " next ( adjoining ant / agent ) , i.e. communicating

 Then becoming " free " again to search for next resume for which it is programmed

A resume may pass from one agent to another ( may be several hundred / thousand ) , till it finally gets received by the Ant / Agent who is " programmed " to find it in the first place !

So , it is not necessary that an Agent Ant finds only that resume for which it is programmed as long as it finds any ONE resume any where & COMMUNICATES ( passes on to the next / adjacent Ant )

Like packet switched network ?

All packets getting assembled by the DESTINATION ANT

Parameters stored getting matched with arriving resume's parameters


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

Page   143

In GuruGem ( improved HARVESTER ) , we are trying to find web - records containing,

>  Company Name

>  Designation ,

but from a single " source " , viz: Google

Our S / W agent " travels " to Google , finds & brings back the results ( food )

But , could we possibly design / devise , millions of S / W Agents , each " programmed " to roam the web ( or predetermined URLs ) and find,

>   One of the thousands of " Designations "  ,  OR

>   One of the thousands of " Company Names "  ,  OR

>   One of the lakhs of " Executive Names " ,  OR

>   Any combination of the above ,

AND,

then bring back the results ?

*   Since most job sites permit access to Job Advt Database to any visitor , without need of a Password , it may be
much easier for Ant Agents to roam Job sites ( like search engine spiders ? ) and bring home suitable job advts ( what we do , in a limited manner )


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

Page   165

Game               =   Online Recruitment / Searching for Employers / Searching for Candidates

Characters       =  >  Recruiters  >  Jobseekers 
 
Agents            =  >  Job Advts   >  Resumes

Environment   =   Virtual  Job  Fair  /  Virtual  Employment  Exchange  (  read my notes )

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

Page     207

I believe Cyril used such an algorithm , to read a plain text resume

The goal was to pick out the " Address " from a plain text resume

The software did manage to accurately find the " Address " ( from anywhere in the resume ) , after about 72 hours of continuous " Exploring / Processing / Learning "

This was nearly 8 years ago !

I am sure , by now , far more powerful Neural Net " Shells " can be freely downloaded , which would parse a plain text resume to find accurately , ALL the fields / values, which we need to create a structured database ( 27 Jan 2007 )
It is simply a question of experimenting . 

And now , we have no shortage of hardware

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

Page    230

Someday , ( when we have a million resumes on our web site ), we will have many " sub sets " of candidates who started their careers ( first job ) , at,

>   same  AGE,  and with

>   same  EDU Quali

Question is :

How did their careers " evolve " ?

What > Salary  > Designation, did each of them reach / achieve , after 5 years / 10 years / 15 years , of experience ?

Was there a " pattern " ?

Did their career paths run " parallel " or did they diverge ?

If the paths did diverge , what other " factors " ( eg; Employer Companies ? ) influenced such divergence ?

Did some plateau out after a time , whereas others continued to climb the Desig / Salary ladder ?

{  see my hand-drawn graph on this page for better grasp }


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

Page    300

*   With an ever expanding set of " Rules " , it will be possible for a " Rule based " system , to delete faked / forged resumes , where a candidate is telling a lie re: some " facts "

*   I have listed several such rules . See my notes on " Expert System " - also my handwritten notes in the margin of book " Expert Systems "

*  In each resume , each and every field value is a " fact "

Therefore , " Facts " are , Age / Exp / Desig Level / Edu Quali / City / Skill keywords ( found in a resume ) / Salary / Company Name ( Employer )


*  One can start with simple rules such as ,

>   " Experience ( years ) "  cannot exceed  " Age "

>   " Age " cannot be less than " 16 "

>   " Edu level " cannot be less than " 10th Std / SSC "

>   "  MD ( Managing Director ) " cannot be less than " GM ( General Manager ) "

>  "  Post Graduate "  cannot be less than  " Graduate "

>  " Salary "  cannot be less than " Rs 1,000 per month "


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

Page   335

In developing Function Profile graphs , we are simply depending on " presence " or " absence " of any given keyword ( Binary State ) . Then assigning " Weightage " { of a given amount or a weightage of ZERO } , depending upon presence or absence

But , in a real world, it may turn out that the candidate with a lower " Raw Score " ( because of many " absent " keywords ) , turns out to be a better choice ( higher interview score in IIT ? ) , than a candidate having higher raw score ( where most keywords were present )

In such a scenario , should we try " FUZZY LOGIC " algorithm ?


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

Page  352 { Chapter on NATURAL  LANGUAGE  PROCESSING }

" Dragon Naturally Speaking " s/w package

Read my notes : ARDIS / ARGIS


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

Page   372

Beysian spam filter ?

based on a starting database of " unwanted " keywords in the email messages


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

Page    376

We must experiment with Kurzweil's paraphrasing software

Could Ray Kurzweil's " Paraphrasing Software " , be based on this ?

See web site of Kurzweil .

Using this , can we " rewrite " a resume ( like creating a step-brother ) from a given " Sample Resume " ?

If such " paraphrasing / re-writing " of resumes can be done online ( automatically ) on our web site , then we could add one more element of FUN - even if the rewritten resume contains some absurd text !

In fact , such absurdity may lend an element of FUN !

Author's Words :  The source for parsing text into bigram chains and then building sentences from them , is surprisingly simple . The bigram is implemented as a simple two-dimensional array . Each dimension is represented by unique words that were parsed from the text ]


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

Page    385

Read page 376

If we succeed in paraphrasing / rewriting a text resume online , we could generate / develop a " Subliminal / Subconscious " function profile graph ! ( - since keywords would have got changed )

This could be fantastic !


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

Page    388

Eg : Agents of jobseekers negotiating salary / terms etc with agents of recruiters

Virtual Job Fair will be / can be " Negotiating Platform " for ,

>  Buyer's Agents  (  Recruiters )

>  Seller's Agents  (  Jobseekers )


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

Page    389

Only yesterday we discussed that in , " Post Job " form , recruiter will add :

" Put into my folder , all future / incoming resumes , having percentile score of > ( greater than ) XYZ  "

Now , he has created an Agent which checks incoming resumes daily & puts into folder " Resumes of Interest "

Also , an email will go out to the recruiter concerned ( alert )

*  Like our GuruGrab ?

[ Author's words : Given mobility , an Agent could be despatched to the remote database to automatically filter the results and then return only what was required by the end user ]

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

Page    440

A few weeks back, there was a report of a mechanical spider-legged robot , which learned on its own to change its " gait " when one of the legs was broken !


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

Page   441

May be we can download if useful in our narrow domain of Online Recruitment

[ Author's words : Cycorp has recently released OpenCyc , which is an open-source version of the CYC technology.

OpenCyc includes a knowledge base ( 6,000 concepts with 60,000 assertions ), the CYC inference engine , and a number of language bindings and APIs to support software development with knowledge base ]


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







No comments:

Post a Comment