Name Match V2

Introduction

The Name Match API is designed to compare two given names and provide a comprehensive analysis of their similarity. This API can be used in various applications, including customer onboarding, identity verification, and data deduplication, where accurately matching names is critical.

Sample Curl

curl --location 'https://api.signzy.app/api/v3/matchers/nameMatchV2' \
--header 'accept: application/json' \
--header 'Content-Type: application/json' \
--header 'Authorization: <Token>' \
--data '{
    "nameBlockv2": {
        "name1": "<Name 1>",
        "name2": "<Name 2>"
    }
}'

Input Parameters

PARAMETER	DESCRIPTION
Content-Type	Contains the id parameter returned from the login step
Authorization	application/json
nameBlockv2	Contains names to be matched
name1	First name to be matched
name2	Second name to be matched

Sample Response

{
    "result": {
        "name1_vs_name2_matchScore": 1,
        "name1_vs_name2_matchResult": "Direct match",
        "name1_vs_name2_matchReason": "The given names are identical"
    }
}

Response Parameters

PARAMETER NAME	DESCRIPTION
name1_vs_name2_matchScore	Name Match score between 0 to 1
name1_vs_name2_matchResult	Direct Match/ Partial Match/ No Match (when task given is nameMatch) Direct Match/Good Partial Match/Moderate Partial Match/Poor Partial Match/No Match (when task given is nameMatchV2)***(refer to the categoriztion table (Table 1) given below)
name1_vs_name2_matchReason	Message output for explaining the reason for the score as a means of text based feedback on what happens inside the code

Details on Output parameters

matchResult

Gives a single class denoting how well the names match. These classes are a grouped representation of a range of scores as categorized below.

Table 1: Score Categorization

MATCHRESULT (V1)	MATCHRESULT (V2)	MATCHSCORE RANGE
Direct Match	Direct Match	1
Partial Match	Good Partial Match	0.85 - 0.99
	Moderate Partial Match	0.60 - 0.84
	Poor Partial Match	0.34 - 0.59
	No Match	0 - 0.33
No Match	No Match	0

name1_vs_name2_matchScore

Gives the raw matching score as a number between 0 to 1. A score of 0 means totally different names and a score of 1 means exactly the same names.

name1_vs_name2_matchReason

Gives a transparent reason why the Name Match model thought the input names are similar or different. This is directly inferred from the modular structure of our algorithm.

*This feature is still under development and we are working on improving it further.

Name 1	Name 2	matchResult	matchScore
RAVINDRA PRATAP SINGH	RAVINDRA PRATAP SINGH	Direct match	1.00
These names are exactly same, so we get a perfect score
RAVINDRA PRATAP SINGH	RAVINDRA PRATAP	Good partial match	0.85
Here, one name (SINGH) is missing but the rest of the names match perfectly. So we are pretty sure it"s the same person
RAVINDRA PRATAP SINGH	RAVINDRA PRATAP S	Good partial match	0.85
Same as above, here we also have an initial “S” for the missing word, so it"s a good partial match
RAVINDRA PRATAP SINGH	R P SINGH	Moderate partial match	0.70
Here, 2 long names are missing (RAVINDRA and PRATAP) so the score is lower but since the initials match, we are moderately sure it"s the same person
RAVINDRA PRATAP SINGH	RAVINDRA	Moderate partial match	0.70
Same as above, 2 names are missing
RAVINDRA PRATAP SINGH	SINGH RAVINDRA PRATAP	Good partial match	1
Here all three names are present in both but in a different order. But overall it"s a good match, we are not penalizing this currently
RAVINDRA PRATAP SINGH	RAVINDRA PRATAP SINGH S O DH	Good partial match	0.90
Here, there are a few extra words (S O DH) and Name Match recognizes it stands for caregiver name (as in Son of DH) and so doesn"t penalize it highly. Overall result is a good match
RAVINDRA PRATAP SINGH	RAVINDRA T SINGH	Poor partial match	0.55
Here, although 2 names are similar but the third initial “T” is completely different which means there is a chance it is a different person so Name Match penalizes it higher
RAVINDRA PRATAP SINGH	RAVINDRA THAKUR	Poor partial match	0.38
Here, “Thakur” is completely different than “Pratap” and “Singh” which means it is highly likely this is a different person, so the final score is very low
RAVINDRA PRATAP SINGH	NARENDRA SINGH	Moderate partial match	0.61
Here, “Singh” completely matches in both names and “NDRA” from both names matches which increases the score but overall it is still a partial match
SHINDE RAHUL MARUTI	S R MEDICO	Poor partial match	0.41
Here, the names are completely different. However, S and R match with the initials of SHINDE and RAHUL so there is some score given. Overall it"s still a poor match
LALAN PRASAD	MAHTO RAMBHAJAN LALANPRASAD	No match	0.20
Here, as humans we can probably see that there is some chance that these two names are of the same person because of the part “LALANPRASAD” in the names but Name Match penalizes it highly due to missing space and words
ATULKUMAR MELABHAI PARMAR	ATULBHAIMELABHAIPARM	No match	0.14
Although we can see there are parts of this name which tells us this might belong to the same person, Name Match penalizes it highly.

Version History

Similarity between two names was calculated by using distance algorithms which did not care for subtle nuances in Names and treated all variations in alphabets, characters, names equally.

V2 Changes

nameMatchV2 runs the new improved version which enacts the improvements described below

Better score categorization

V1 only had "Direct Match", "Partial Match" and "No Match".

V2 breaks it up into:

Direct Match

Good Partial Match

Moderate Partial Match

Poor Partial Match

No Match

This enables more granular score categorization. For example, let's see 2 different cases and the difference between V1 and V2 response

Input: 'Tonmoy Sahu' and 'Aditya Sahu'

	V1 response	V2 response
name1_vs_name2_matchScore	0.50	0.60
name1_vs_name2_matchResult	Partial Match	Moderate partial match

Input: 'Tonmo Sahu' and 'Tonmoy Sahu'

	V1 response	V2 response
name1_vs_name2_matchScore	0.92	0.80
name1_vs_name2_matchResult	Partial Match	Good partial match

Modules to captures nuances of Names

We have introduced modules that take care of groups of user scenarios as shown below. These modules also enable us to return specific matchReason based on the module it hits and give appropriate penalties based on the scenario.

Scenario	Module
Exact names from two sources	Exact match
One of the names is reversed and many times has a comma	Sequence
Users name in One ID has a middle name the other doesn't	Word missing
Users name in One ID has initial of one name the other doesn't (Tonmoy J Borah - Tonmoy Borah)	Initials
The OCR of one ID card has removed spaces b/w a name hence the name from one ID card has no spaces	Space
The user's name in one ID card has a different spelling but the same pronunciation as some names in Hindi and English are spelled differently. Like RAJEEV and RAJIV	Spelling substitution
Names are taken from an ID with Salutations Like Mr and Dr	Pretreatment
Two names are Naval Kishore and NKishore or some letters are missing	Subset
Similar spellings of names (agarwala - agarwal, mahammad- mohd) in certain cultures	Common Word Substitution
Sometimes names have caregiver names (eg. Aditya Jain S/O Ramesh and Aditya Jain)	Care of module

We plan to continuously improve and add support for new customer expectations and real world scenarios which will be bundled into different modules and either be included into the Default version or be kept separate.

Status Codes

CODE	MESSAGE
400	Error: Name cannot be blank or spaces or anything except string
500	Internal Server Error