By - Munish Moudgil IAS, Commissioner, Survey Settlement and Land Records, Revenue Department, Government of Karnataka

Artificial Intelligence, trivially, is the “Intelligence exhibited by machines”. We intuitively find a machine intelligent when it “behaves” like us. Alan Turing, the father of Computer Science and AI, prescribed the same test to call a computer “intelligent”.

While the Government already has the information with respect to a citizen to instantly decide/deliver the request she is asking yet for eons she waits as this information is in silos and Government cannot use it. First we ask her to bring it (from us!); and once she does we don’t believe it as she brought it. So we set out to verify it while she waits!! An omniscient person with perfect recall, with all above information, would revolutionize Citizen-Service Delivery by immediately deciding/delivering the request. The way to do it is by identifying a given person in different databases and pulling out relevant information the moment person makes a request to the Government

This hasn’t yet been achieved.

Aadhaar attempts to do it but the process is excruciatingly slow as Aadhaar of each of millions of persons have to be entered manually against each name in each database separately AFTER field verification. This will go on for years if not decade. And even when achieved it does not solve the problem fully as explained in paragraphs below.

If one picked, say, Ration Card database of a village and actually compared with MGNREGA Job Card database of the same village, it is obvious that one would end-up finding at least 70% names in MGNREGA Job Cards among the ration card names.

While comparing like this is manually impossible even computer softwares have not performed well wherever such an attempt has been made. Further, any “intelligent” software would cost crores of rupees and its efficacy isn’t assured.

In a parallel problem, which government faces every day, if Munish approaches government and asks that a house be sanctioned to him as he is houseless; it is next to impossible to verify whether or not in the past he ever had got a house from government. Aadhaar seeding route does not help as people won’t come forward to give information to include their aadhaar numbers against benefits they have already availed in the past for which the information lies in old databases.

Name Matching Algorithm is a software programme created and written by the undersigned to precisely solve these problems. Simply put – given a “name” in 1st-database/list it identifies all names in 2nd -database/list which “match” it. Further, it returns percentage “match”; enabling one to pick the “best match” or the “acceptable match”. For example, if “Munish” and “Manish” are compared the software will return MatchScore of 93.5%. While MatchScore between “Munish” and “Mukesh” is 79%, and, that between “Munish” and “Sukesh” is 35%. This way, depending upon user requirement, the declaration of “match” or success, could be set at any threshold of MatchScore.

In declaring a “match” the algorithm closely mimics humans in the sense that given two lists, whichever names would be declared a match by a human, the algorithm is likely to find and declare each of them as “match”. As different persons and different contexts could require different yardsticks, thus, threshold for declaring a “match” is in hands of the user. In this, the 100% “match” threshold would mean that only perfectly identical names would be declared a “match” while at a lower percentage the “match” would be declared more closely resembling human behavior

Present usage of NameMatchingAlgorithm in Karnataka:

1. Karnataka remitted Aadhaar (AEPS) and ACH-DBT based payments to about 40 Lakh farmers for crop loss in 2016-17 using NameMatchingAlgorithm as follows

a. Field Revenue Officials (Talati/Patwari/Village Accountant) submitted crop loss field report with farmer name and survey numbers.

b. NameMatchingAlgorithm compares farmer name in VA report with the farmer name in Record of Rights (RTC) (Bhoomi database). Those field farmer names which didn’t match owner names in the RTC were weeded out.

c. Then NameMatchingAlgorithm compared farmer name in RoR/RTC with farmer’s name in Aadhaar. Again, the names which didn’t match were removed.

d. Finally, NameMatchingAlgorithm compared Aadhaar names with the bank account holder name with the said Aadhaar Number (even when Aadhaar seeding hasn’t happened the name-matching allows to compare Aadhaar name with Bank Ac name thereby effectively achieving purpose of Aadhaar seeding). Here again if Bank Ac Holder name didn’t match with Aadhaar Name; the Aadhaar or ACH-DBT based payment was held back.

Finally, after due verification more than Rs 2500 Crores to about 40 Lakh farmers paid.

2. Over The Counter (OTC) Issuance of Caste/Income/Residence Certificates:

a. Digitized certificates are being issued for more than 10 years in Karnataka, and, large database exists of certificates already issued.

b. Further, a door-to-door enquiry has been done for full population, and, now, caste, income and residence certificates are pre-created and kept ready for more than 5 crores.

c. Now, the moment citizen approaches front end of common service centre and gives his name; the NameMatchingAlgorithm fetches existing certiicates for that name in the said village. The citizen can select his certificate and print and take on the spot.

3. Housing Programmes:

Any new beneficiary wanting a subsidized house, his/her name is compared with all beneficiary names and guardian names using NameMatchingAlgorithm, and, in case of “match”, the new beneficiary is declared Name-Duplicate.

It can be appreciated that uses of Name Matching Algorithm are huge. The SDK of the NameMatchingAlgorithm developed using Microsoft .Net framework can be downloaded here and may be used by the government departments free of cost, with due acknowledgement to the author of the algorithm.