Monday, July 27, 2020

The NAME search problem

There are countably finite number of people and organizations with similar names. Any computerized system attempting to find the exact person (or entity) is extremely difficult to build. 

There is an other angle to the problem, as well. The number of (identity) cards a person carries these days is pretty high.

1. Aadhar Card
2. PAN Card
3. Driver's License
4. Voter ID Card
5. Credit Card
6. Debit Card
7. Ration Card
8. Passport
9. Mobile Phone (registered with the telco)

There might be more that I may have missed. What happens is that in rural India, the same person has different names in these different cards which opens up the system for abuse. My building caretaker is called 'Shambhu'; phonetically this is the exact spelling but he happens to have different names such as 'Sambhu' or 'Shamvu' in different cards. Recently money came to his Jan Dhan account and got automatically deducted for a gas cylinder that he never received. 

Prime Minister Modi's ambitious plan of linking everything with Aadhar could have solved it but Aadhar has its own problems

In search engines precision and recall are inversely proportional. This problem cannot be solved with search engines like Solr or ElasticSearch.

The only name search that works perfectly are Domain Name Systems. Although domain name servers are examples of the best distributed databases, the effort is centrally co-ordinated by ICAAN. A politically fragmented world can probably never be able to solve this pressing issue of true identity. 

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.