Fuzzy Match Names

Available lang-locales: This service is configured for en-US. Note that it can also be used for other languages, such as de-DE, but the graphemic-phonetic mapping may be slightly different in other languages.

The Fuzzy Match Service (FMS) is designed to enhance caller experiences by significantly improving the accuracy and efficiency of name verification processes. It adeptly handles simple spelling variations (such as, "Jon" vs. "John", "Annike" vs. "Annika"), which often challenge basic text-matching systems. By leveraging advanced matching algorithms, the FMS ensures minor discrepancies in name transcription do not obstruct the verification process. This service is invaluable in scenarios demanding high precision in name matching or verification, such as during user authentication in customer support calls or product name confirmation in sales interactions.

Technical Configuration

The service is configured within the bot as follows:

Branches

• success: The service was able to process the provided input.

• else: Added by default; indicates a failure of the service.

Input

• originalString: (required, String) The value you want the input to be tested against, i.e. the "correct" string you are comparing to.

• stt: (required, String or Array) User input as a string, or the array of stt_candidates returned by the platform.

• lang: (optional, String) Currently this parameter is optional, as the service is only configured to use graphemic-phonetic mapping for en-US. Input is expected in the 4-character lang-REGION format.

Output

• MatchStatus: Either exact, fuzzy, or not a match

• operations: Only relevant for fuzzy matches; this output contains an object explaining the operations/steps used to get a match e.g. ["Spurious 'h' between nicXolas and nicholas;"]; if not relevant, this will

  • Example: looks like: {"originalString": "Nicolas","Stt": "Nicolas","Ops": {}}

• stt_candidates: If MatchStatus is fuzzy AND stt contains the stt_candidates it will return all the fuzzy matches inside an array; otherwise this will be an empty array.

Technical Configuration

• URL: https://parloaservices.azurewebsites.net/api/FuzzyMatchNames?clientId=<CLIENT_ID>

• Header: x-functions-key: <AUTHCODE>

For authentication, replace <AUTHCODE> and <CLIENTID> with the values provided by your Parloa representative.

Additional Details

Hosting Environment

EU only

Sample CURL

curl --location 'https://parloaservices.azurewebsites.net/api/FuzzyMatchNames?clientId=<CLIENTID>' \
--header 'Content-Type: application/json' \
--header 'x-functions-key: <AUTHCODE>'
--data '{
    "input": {
        "originalString": "Nicolas",
        "stt": [{"text":"Nikolas","confidence":0.5439518},{"text":"nikolas","confidence":0.48069397},{"text":"nicola","confidence":0.3820934},{"text":"Nicolas","confidence":0.47726634}],
        "lang": "EN"
    }
}'

Sample Inputs and Outputs

# Single string, no stt_candidates, exact match:
{
    "input": {
        "originalString": "Nicolas",
        "stt": "Nicolas",
        "lang": "EN"
    }
}

{
    "choice": "success",
    "output": {
        "MatchStatus": "exact",
        "operations": {
            "originalString": "Nicolas",
            "Stt": "Nicolas",
            "Ops": {}
        },
        "stt_candidates": []
    }
}

# Single string, no stt_candidates, fuzzy match:
{
    "input": {
        "originalString": "Nicolas",
        "stt": "Nicholas",
        "lang": "EN"
    }
}

{
    "choice": "success",
    "output": {
        "MatchStatus": "fuzzy",
        "operations": {
            "originalString": "Nicolas",
            "Stt": "Nicholas",
            "Ops": [
                "Spurious 'h' between nicXolas and nicholas;"
            ]
        },
        "stt_candidates": []
    }
}

# Single string, stt_candidates, exact match:
{
    "input": {
        "originalString": "Nicolas",
        "stt": [{"text":"Nikolas","confidence":0.5439518},{"text":"nikolas","confidence":0.48069397},{"text":"nicola","confidence":0.3820934},{"text":"Nicolas","confidence":0.47726634}],
        "lang": "EN"
    }
}

{
    "choice": "success",
    "output": {
        "MatchStatus": "exact",
        "operations": {
            "firstName": "Nicolas",
            "firstNameStt": "Nicolas",
            "MatchStatus": "exact"
        },
        "stt_candidates": []
    }
}

# Single string, stt_candidates, no exact match:
{
    "input": {
        "originalString": "Nicolas",
        "stt": [{"text":"nicolas","confidence":0.5439518},{"text":"nikolas","confidence":0.48069397},{"text":"nicola","confidence":0.3820934},{"text":"nicholas","confidence":0.47726634}],
        "lang": "enUS"
    }
}

{
    "choice": "success",
    "output": {
        "MatchStatus": "fuzzy",
        "operations": {
            "originalString": "Nicolas",
            "Stt": "nicholas",
            "MatchStatus": "fuzzy",
            "operations": [
                "Spurious 'h' between nicXolas and nicholas;"
            ]
        },
        "stt_candidates": [
            [
                {
                    "originalString": "Nicolas",
                    "Stt": "nicolas",
                    "MatchStatus": "fuzzy",
                    "operations": [
                        "all differences are vowel-only;"
                    ]
                },
                {
                    "originalString": "Nicolas",
                    "Stt": "nikolas",
                    "MatchStatus": "fuzzy",
                    "operations": [
                        "corresponding c:k;"
                    ]
                },
                {
                    "originalString": "Nicolas",
                    "Stt": "nicholas",
                    "MatchStatus": "fuzzy",
                    "operations": [
                        "Spurious 'h' between nicXolas and nicholas;"
                    ]
                }
            ]
        ]
    }
}

Usage in Parloa

Sample configurations with Output

Configuration with Input String

The correct name of the user as obtained from the database is provided to the service here via the variable firstName_stt.

The user input is provided here via the variable firstName_stt.

Output: Fuzzy Match

In this case, the system transcribed Annika and the service was able to fuzzy-match that to the first name Anike provided from the database.

Output: Not a Match

In this case, the system transcribed Monika; the service was not able to match that to the first name Anike provided from the database due to the additional consonant M at the start of the word. For situations like this, it would make sense to design the dialogue with a second question or a fallback option in case of failed validation.

Important Points to Note

Ensure to handle all customer data with utmost care, adhering to data privacy regulations. Additionally, always verify that CRM or other service lookups are functioning correctly to fetch accurate customer details.

• Our service maps letters to their sounds, focusing on common phonetic (sound) patterns. This approach is highly effective but can be challenging for languages without direct spelling-to-sound correspondence.

• Differences in vowel sounds alone are treated as fuzzy matches by our service (such as "Multi" vs. "Malta"), reflecting our design to accommodate variations across dialects and second-language pronunciation. This can be seen in the operations output parameter, where in such cases, it will note the operation 'all differences are vowel-only'.

Decision-Making Criteria

When deciding on match acceptance:

Exact Match

Proceed with confidence.

Fuzzy Match

Use Operations details to determine if the match is acceptable based on context.

No Match

Re-confirm the details with the customer.