Meet the code, coder behind crop loan waiver software

A unique algorithm written by an IAS officer is insulating lakhs of farmers from disqualification under Chief Minister H D Kumaraswamy’s ambitious crop loan waiver scheme as their names in their documents are not matching with bank records.

Over 12 lakh farmers have submitted Aadhaar cards, ration cards and land details to commercial banks so far to get their crop loans waived. About 2.87 lakh were disqualified as their name in the loan accounts were not the same as those in their documents. Thanks to the algorithm, over 1.44 lakh farmers have qualified again and others might, too.

The name matching algorithm developed by Survey, Settlement and Land Records Commissioner Munish Moudgil is generating matching score for names in a matter of seconds, allowing authorities to expedite verification of crop loan data and identify genuine farmers. This negates the need for names to be checked manually, a laborious task. It also helps check fake loan waiver applicants.

“We’ve now built in artificial intelligence into the system so that name-matching happens akin to how a human would do it,” Moudgil, said. The IIT Bombay alumnus said he volunteered to develop the software to implement the Rs 46,000-crore loan waiver.

The algorithm looks for a match when a name between two databases is not the same - they could be spelt differently, contain initials/surnames or, worse, are written in different languages.

“For example, if ‘Munish’ and ‘Manish’ are compared, the software will derive a match score of 93.5%. But with ‘Manish’ and ‘Sunish’ the score will be a little over 50%. That’s the kind of intelligence built in,” he explained.

“I’ve written the algorithm such that the programme selects a string of best matches and comes out with a name-match score to determine whether it is the same
person.”

The name-match threshold has been set at 60%, which means authentication of a farmer whose name does not achieve a match score of over 60% will be considered as failed. Such applications are referred to taluk-level committees for manual verification.

Moudgil first wrote the algorithm in 2014. It has since been used by the Rajiv Gandhi Rural Housing Corporation Limited and for the Parihara drought relief
compensation.

“Maharashtra’s loan waiver wasn’t successful because their authentication wasn’t fool-proof,” Moudgil said. “Say, an Aadhaar number does not change, but the person using it may change (duplication). The conventional one-time-password (OTP) authentication didn’t work in Maharashtra as the registered mobile numbers were no longer in use. Hence, authentication became a problem.”

The crop loan waiver scheme software architected by Moudgil has integrated bank account data with databases of the Unique Identification Authority of India (Aadhaar), the Food & Civil Supplies department (ration card) and Bhoomi (land records).

Filtering out fakes

“If farmers have loans in both cooperative and commercial banks, only their cooperative loan will be waived. The Aadhaar, ration card and land details result in sanitising. We project that the government will save about Rs 6,000 crore by identifying duplicates,” Moudgil said.