IMPROVING EXTRACTION ACCURACY

Kartavya

Resumix advises American jobseekers to format their resumes under following broad/common “headings”:

1. Objective

2. Experience

3. Employment

4. Work History (sub-group / synonyms?)

5. Positions Held

6. Appointments

7. Skills

8. Summary

9. (Summary of) Qualifications

10. Accomplishments

11. Strengths

12. Education

13. Affiliations

14. Professional Affiliations

15. Publications

16. Papers (sub-group / synonyms?)

17. Licenses

18. Certifications

19. Examinations

20. Honors

21. Personal

22. Additional

23. Misc

24. References

---

Not every American jobseeker would use all—for that matter no jobseeker may use all of these “headings”! Every American jobseeker may use some combination of these headings.

Our American version of RESUMINE would need to recognise these headings & break-up a given resume into several paragraphs before it attempts to “extract” values for specific fields in each paragraph.

We should get a good working American version of RESUMINE ready before Nirmit leaves for Brussels in May.

✍️

25/02/03

---

(22/03/03)

Improving extraction accuracy

Abhi told me last evening that he has planned to process/convert 1000 email resumes today morning.

Based on this “experiment”, remaining 26000 can be processed next week. When this is done, a tabulation such as ANNEX A should be prepared. It can also be rearranged in the descending order of (2nd column) No. of Failure Cases out of 27000. Descending order will tell us, which “cases” to tackle first (priorities).

You may also consider constructing a TOOL screen, as shown in ANNEX B. I have slightly modified existing Resumix screen. Several human experts could be asked to work on this tool simultaneously (different “failure” cases may be assigned to different experts). After studying each “failure-field”, these experts should enter their comments on the middle block.

Of course, to speed up the process, Vithal/Santa may go thru all “failure cases” in advance & simply forward those cases for which software value does exist but the software failed to extract the same.

With this tool, we will be able to capture the knowledge of several experts.

TEMPORARY REGISTRATION FORM

(FOR FREE DEMO SERVICE)

Dear Corporate,

By filling in this form, you can avail of our webservice free-of-cost, for a period of one month from date of registration or extracting data & creating a structured database for 500 resumes – whichever is early.

Immediately upon registration, you will be allotted ONLINE:

• a temporary Member No.

• a User ID

• a Password

You will need to use/enter these every time you login for service.

At the end of the free demo service, should you decide to avail of this service on an ongoing basis, you will need to fill in, online, a PERMANENT REGISTRATION FORM.

Entering your temporary Member no/User ID/Password, in that Permanent Form, will enable you to enter additional details required by that Form. When you do that, you will be allotted online, a Permanent Member No. At that time, you will also be able to enter/select any no. of User IDs/Passwords, depending upon the no. of persons in your Organisation authorised to use this webservice.

TEMPORARY REGISTRATION FORM

(For Free Demo Service)

Name of Company:

Address:

City / PIN (ZIP):

State (In case of Indian Companies):

Country:

E-mail:

Phone: 1, 2, 3, 4

URL:

Name of Executive Registering:

• Last Name

• First Name

Designation of Executive Registering:

Phone Nos:

• Direct No.

• Mobile No.

• Resi No.

• Email (Personal):

Products / Services offered by the Company:

---

[Submit]

Message appearing upon clicking:

Dear [Name],

Thank you for registering. Please note down following for login:

• Temporary Member No.: [ ]

• User ID: [ ]

• Password: [ ]

We should be able to add/delete thru ADMIN TOOL.

AWARENESS

We were made aware/recommended your webservice by:

(You can tick only one source. In case you do not tick any, it will be assumed that you chose to use our webservice on your own or upon recommendation of persons/entities not listed below):

Srl. No.	Name of Organization which recommended to you	Tick one
1		○
2		○
3		○

In Case of a Person, type

• Name: Last Name [] First Name []

• E-Mail: [_____]

• Resi. Phone: [_____]

---

ACCEPTANCE

• We are happy to welcome you to 3P Webservice.

• Your Permanent Member No: [_________]

• Your User IDs & Passwords will be as per details entered by you in this form. Kindly inform these executives that now they can log-in using their own IDs/Passwords. Make sure each executive is given only one User ID/Password combination. Of course, all of these executives will need to log-in using same Permanent Member No. allotted above.

By using his/her own unique User ID/Password, individual executives will be able to view INDI. STATISTICS OF TRANSACTIONS which only he/she has carried out.

He/She will not be able to view transactions carried out by any other executive.

Consolidated statistics of transactions are viewable from CONSOLIDATED STATISTICS OF TRANSACTIONS.

However, these can be viewed only by the executive whose name appears FIRST (on top) in the AUTHORISATION LIST page.

The PREPAID Service Fee will stand reduced by the “Per Transaction Tariff” (see TARIFF), on a continuous/automatic basis, as each transaction is carried out. An auto email reminder will be sent when only 10% of “Service Fee” amount is left balance in your account. To avoid disruption in service, always ensure to make further payments before the balance stands reduced to “NIL.”

Kartavya

15/02/03

Functional Specifications

In my yesterday’s note, I have mentioned about 3 Screens viz.:

• DEMO Screen

• EDIT Screen

• VIEW Screen

I also talked about constructing these in a MODULAR way, so that we could have several building-blocks (Screens), which we can re-combine as we want, to create any desired Screen.

Enclosed find my suggestion on how we could try to achieve this.

You will notice that I have added one more screen called “Rating Page.” This will form part of VIEW Screen when a HR manager:

→ Conducts a “Search”

→ Sees short display of search results

→ Clicks on a particular resume against name of any particular candidate (record) in the resume (—so we can make money!)

→ Decides to email the VIEW Screen to a colleague (or several colleagues) for reading/assessing/filling the Rating Page & then emailing (—so we make some more money!!)

OR

→ Finds missing “fields” & decides to email to concerned candidate (with a covering note), asking him to fill-in a return (if he wishes to be considered for this position).

(—So, we continue to make money, every time email button is clicked!)

OR

→ Finds everything (all fields) are in order/complete, but finds that the extracted resume is 3 YEARS OLD (We should highlight the “DATE” somewhere so that the obsoleteness of the resume becomes bold & clear) and therefore the HR manager wants the concerned candidate to UPDATE/EDIT it. So he once again “clicks” email (of course with a provision for a STANDARD covering letter).

(—Some more money!).

Whereas,

We may charge only Rs. 10 for extraction, we should earn Rs. 20 for “Service.”

We should target to earn Rs. 20/- per resume from the clicking of EMAIL button!

(even @ Rs. 2/- per click!!).

Second thing which I now propose, is to remove/retain COMPETENCE-PROFILE from the EDIT Screen.

In my yesterday’s note, I said, even this Competence Profile will be part of EDIT Screen & get emailed to concerned jobseeker.

Now, I would like to reserve this feature to only our own website.

That is,

• When a jobseeker directly submits his plain text email resume on our website, he can see/download/printout his Competency Profile.

But he will NOT get his Competency Profile in the following cases:

(a) When one of our subscriber companies sends (emails) EDIT Screen to him for updating & returning.

(b) When we convert (i.e. extract) & upload all the 80,000 email resumes (that we have accumulated over the years) on our website, and the concerned email goes out to the concerned candidates with the EDIT Screen.

(We may give him a hint that had he submitted his resume DIRECTLY on our website he could have also got his Competency-Profile!)

Of course this (approach) throws-up a possibility that all those 80,000 (+2000 which we hope to receive daily under our Project Manhattan) may ONCE AGAIN resubmit their simple/plain text email resumes directly on our website! – in order to get a copy of their Competency-Profile.

There could be a huge amount of duplication.

Of course, no one knows for sure, what can/will happen with each web-initiative. We would need to feel our way as we go along, and be ready to change course, at short notice.

Kartavya

RESUMINE / RE-SEARCH

WEB SERVICE / PHASE I

14-02-2003

FUNCTIONAL SPECIFICATIONS

---

Extraction / Mining

From unstructured email resumes, RESUMINE should be capable of extracting specified information and create a structured "Database", which lends itself to search, using various "Fields" as "Search-Parameters".

Info. to be extracted will also contain important / critical / frequently occurring keywords. It will also extract "least frequently occurring" keywords (for the purpose of Profile Enhancer).

---

Database Search

Webservice must have a "Search-Engine" which can search the structured database (created thru RESUMINE), using a wide variety of "Search-Parameters". The search could be sequential or parallel.

There should be a "Refined Search" feature.

It should also be possible to search the database, using keywords / words (text).

Competence Profile

The software should be capable of generating graphical "Competence Profiles" based on:

• List of predefined functions.

• List of important keywords appearing in each function.

• Probabilities of occurrence of these keywords, in an incoming resume.

---

Display

The software should be able to display as per following Screens (UI):

• Demo Screen

• Edit-cum-ARGIS Screen

• View / Display Screen

---

Demo Screen

This screen will be used for demo purpose only.

In left-hand vertical will be "fields-names". The fields which cannot be captured, will light up in "RED", giving clear indication of failure.

For rest of the fields, where...

Data/value has been successfully captured, moving the cursor on any field will turn into red (or some other method of highlighting) the corresponding value in the body of the resume. Perhaps the two can be connected with a red line.

Keywords extracted & displayed in the right-hand column

• Shall not exceed 20

• Shall get arranged in the same order in which these occur in the body of the resume.

From top-to-bottom, the keywords shall fill up the vertical bar, one-at-a-time, in a sequence, as each keyword is “snatched” from the body of the resume.

Moving the cursor on any keyword shall “light-up/redden” the corresponding value in the body of the resume.

Right-hand part of the DEMO screen will display the competence profile bars horizontally.

• Name of function should appear within (inside) the bar.

• % (percentage) value should appear on the right-hand of each bar in a different colour.

A maximum of 5 competence bars…

…may be accommodated (in which case, we may not need a vertical scrolling feature). This will simplify matters.

The bottom of the right-side screen (below Competence Profiles) shall display “Rarely / Least frequently occurring keywords” found in that resume.

Resume body (text) itself may be too long to be accommodated within the screen, so it will be necessary to provide a scrolling bar.

When you move the cursor on any particular keyword (in the Keywords Extracted column), the body of the resume shall automatically move / slide up-or-down, and bring that keyword’s corresponding value (text) in the body within the view. (Of course, it should also turn red at same time).

Such a horizontal “aligning” of the words (in text-in column) and values (in body) automatically will make it very easy for the viewer, reducing his search effort/time.

Different colours should be added to the DEMO screen without making it appear “Gaudy/Cheap/Garish.”

DEMO screen should carry several “buttons” at the bottom, indicating a variety of “transactions” that can be…

…achieved/accomplished.

If possible, DEMO Screen should carry 3P Logo somewhere.

---

“EDIT-CUM-ARGIS” Screen

In this screen all the data/fields which have been extracted/not extracted would appear in a sequence of “Blocks.”

Each such “Block” will group together fields which have a common “theme/subject.”

Each block should have a clear/bold “TITLE” in capital letters.

Essentially this screen is to be used for the following purposes:

(A) For editing of any value by the recruiter himself – in case software fails to capture such value.

(B) For recruiter to forward this screen (as an email) to the concerned jobseeker along with appropriate covering letter, asking the jobseeker to himself “edit it” & return email it back to the recruiter.

This process will give a chance/opportunity to the jobseeker concerned to make his profile complete/comprehensive by finding out what fields the recruiting company requires and fill those in, in case missing – via return email.

So the ball is put back into the jobseeker’s court!

If he wants to be “considered” for a particular vacancy advertised by the company, then he better take the trouble to add it in our structured database (including filling-up of missing values/missing fields).

This will reduce/eliminate the “edit-work” from our subscriber company & transfer to the jobseeker. Recruiters would simply love this feature! And this should happen at the click of a button (– for which, of course, we will charge!).

This screen should also carry 3P Logo.

A jobseeker, after “editing” this screen – and after inserting missing values – should be able to:

→ Email it to any other company

→ Email it to any other Placement Agency

→ Take a printout

That means, this (ARGIS) Screen should also carry “COMPETENCY PROFILE” (graphically).

---

VIEW/DISPLAY SCREEN

This Screen will popup when a recruiter clicks on VIEW button against the name / PEN of any candidate whose data appears in the

SEARCH RESULTS

PEN	Name	Edu	Age	Exp	Ind	Funct	Click VIEW	Shortlist
							[VIEW]	[ ]
							[VIEW]	[ ]
							[VIEW]	[ ]

Idea is that the recruiter should see a button to view the detailed resume along with highlighted keywords & competence profiles (graphical display of structured fields) in order to make up his mind whether to shortlist this candidate or not.

So at the bottom of the VIEW/DISPLAY screen, there should be a checklist for shortlisting.

At this stage, the Recruiter may also want to:

• email the “DISPLAY” Screen to a colleague within the organisation for his evaluation/assessment/comment → to return back to him with appropriate ANNOTATIONS.

• take a print-out (or printouts) to give to boss / or to members of interview-panel (who will need while conducting interview).

If subscriber is a Placement Agency (or one of our remote franchisees), he would certainly like to forward such a “Viewed + Shortlisted” resume to his client for a specific research-assignment.

Such a Placement Agency would most certainly like its own Logo to appear on each such email or printouts.

So, there are 3 DISPLAY SCREENS (UIs):

• DEMO Screen

• EDIT

• VIEW

Perhaps we could design these in a MODULAR way & recombine modules to create each screen.

Dated: 10/02/03

Kartavya

RESUMINE DEMO (Beta Version)

• We shall give this demo to L&T Group on March 3rd – possibly afternoon 3 p.m.

• I will get concerned L&T officers to confirm date/time/venue. Most likely venue will be their “C” bldg Conf. Room which has an LED projector hanging from the ceiling & a big screen. The room can conveniently sit 20 people. I will arrange for you/Abhi to visit Powai on 2nd March & carry-out a “dress-rehearsal” of the demo. An engineer from their electrical maintenance dept. will remain present on both the days, to help you carry-out necessary connections.

• We shall carry Nimit’s laptop duly loaded with Resumine software. I will get Mr. Siroor to bring along a CD containing some 50 email resumes. Nimit’s laptop should read this CD, import one resume at-a-time & convert. Do you need to give any special instructions to Mr. Siroor?

• Do you want me to arrange for a colour printer (for attaching to Nimit’s laptop)? It would be a good idea to enable L&T HR managers to take colour print-outs of their own “converted” email resumes as a “proof” of Resumine’s capability.

They (HR Mgrs) may wish to show such colour print-outs to their (Director) Bosses, who may not be able to remain present e.g. A.M. Naik / J.P. Nayak / Y.M. Deosthalee / Mukhija / KV / AR etc.

Of course, I will try to get as many of these to remain present, as possible.

We must try out / thoroughly test this “printing” feature at our end first.

• Should we prepare 2/3 powerpoint slides to precede our demo? These would explain the problems faced by corporates & how Resumine “solves” these problems.

• Demo of RESEARCH (search-engine) should be combined with Resumine demo. What is the use of converting email resumes into a “structured” database, if the database cannot be searched?

Now, when we give Resumine demo, at the most 8/10 resumes may be converted.

This is too small a database for conducting a search!

Therefore, let us (on a separate CD) carry our own 5000 converted email resumes (database) with us, so that we can load this 3P database to give demo of RE-SEARCH.

➡️ It is important that these 5000 are duly converted/extracted using Resumine. This is critical.

Because when we click on any of the PEN in the DISPLAY (shortlist of result), then we should be able to see on the screen the CONVERTED RESUME with “Competency Profile” / keywords & everything!

Pl. feel free to ask.

cc: Abhi

Kartavya / Abhi / Inder

30-01-03

RESUMINE (Resume Extractor)

‘Resumine’ would have following aspects:

Versatility

Is it versatile enough to carry out datamining from the widest variety of email resumes? e.g. not only resumes of Indians (found on Indian job sites) but also of Americans?

Although, initially we may release V 1.0 for only Indian resumes, next version must cover American resumes. After all USA/UK are much bigger markets – and what is more interesting, “better paying” markets! This is crucial if we wish to leverage our association with PENHRYNE to penetrate these markets.

Depth

As of now, we have some 17 “fields” & 6 “blocks”. In Version 2.0, these would need to be enlarged (see my earlier note).

► Accuracy

This is a very important aspect.

We may release V1.0 with 80% accuracy but soon follow it up with V2.0, giving 90% accuracy.

By Jan 2004, we must target 95% accuracy (– balance 5% errors should be capable of being “edited” manually, online).

---

► Dependability

The software must perform reliably again & again & again! Then only can our reputation spread by word-of-mouth (which is our most powerful SALESMAN!). After all, community of HR managers is quite small and word spreads like wildfire – whether “Good Word” or “Bad Word”!

---

Page 3/7

► Presentation / Packaging

No matter how good maybe our “Resumine” in its “functionalities”, if our visual DISPLAY is poor, nobody will want to touch it!

Our User Interface must be totally intuitive and extremely user-friendly. And then it must “dazzle” the viewer. It must have “Magic-like” quality! It must “spell-bind” the viewer and hold him in “awe”! It must “amaze” him!

The display (UI) must prompt user to play-around/interact/explore. It must keep him “glued.”

With this in mind, I ask you to examine enclosed proposal re: its feasibility.

The screen is divided in 2 parts, vertically.

Left side will display actual resume which is being “extracted”.

…and turning red a Value when the cursor is moved across it, is to DRAMATIZE / MESMERIZE the viewer with the

• Elegance

• Speed

• Accuracy of data capture

(In Version 5.0, moving the cursor will make the computer to speak-aloud the word! — Text > Speech Conversion)

As far as the right-hand vertical bar of “keywords extracted” is concerned, initially it will be blank/white space. As soon as match-maker has found/located/matched the keywords, this bar will suddenly get filled-up, in max ONE SECOND!

Here we can even design the software in such a manner that the order in which the keywords get arranged in the “bar” is exactly the same as the order (sequence) in which these keywords appear in the resume, so that the resume does not have to keep dancing up & down all the time.

Once again, moving the cursor on any keyword will turn it red (in the resume) and highlight it.

…the “Fields” (being extracted) and on the other side are the “Keywords” which “match-maker” tool finds in the resume.

The idea is,

• When viewer/user moves the cursor on any given “field”, the actual “value” (word/words/para) in the resume will change colour (say “Red”). Of course if that word/value is hidden at the bottom, entire resume will automatically slide-upward and “expose” the concerned value/word.

The highlighting — and turning red — of the relevant word is the proof that the software has accurately/correctly captured the value!

If desired, we can also provide another screen/s (like the ones developed by Sunitha/Vikram) — and a button so that a “doubting” user can actually see the captured value INSIDE a field-box! — and be able to edit the same.

But the idea of “highlighting…

a simple “sequence,” we are avoiding the 9 “category–boxes,” as also the confusion (difference of opinion) whether a given keyword is:

• Skill

• Knowledge

• Attitude

• Attribute
  etc.

When we come out with Version 2.0, two things will happen, when cursor is moved across a keyword in the vertical bar:

viz:

#1 → Keyword gets highlighted in the resume

(Covered under Version 1.0)

PLUS

#2 → Next to that keyword a box appears containing “synonyms” picked up from MS Thesaurus.

e.g.:

Analyse ───────> Related Concepts
                 • Examine
                 • Study
                 • Observe
                 • Check-out
                 • Dissect
                 • Measure

I presume, integrating MS Thesaurus with highlighted keywords (in the resume), may not be very 
complicated, but you can imagine the impact on the viewer! How Visualthesaurus.com “GRIPS” us!
The entire right-hand half of the screen, I have set aside for displaying the COMPETENCY PROFILE.
We may restrict the display to 5/6 bars.
(With scrolling arrangement, all the bars can be accommodated).
However, from the viewpoint of the recruiter / HR manager, what would be of great interest 
is the “PROFILE ENHANCER” box at the bottom.
This would contain the +rare keywords found in this resume, which (may) set this candidate apart 
from hundreds of others. These rare keywords may contain “clue” to some unique skill / knowledge 
that candidate possesses. Now imagine the (pleasant) surprise of the recruiter, if he is 
looking for exactly such a unique skill!
Please discuss with me soon to finalize the “Integrated” display.
(+ Very low “frequency of occurrence”)