Genesys Cloud - Developer Community!

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Hi Sophie -
I needed a warm-up this morning so I made a little test AWS lambda function that does this. 

If you promise not to use it for prod (laughing) I can give you the URL and config to make it work as well as a temporary token.

Here's the results:
{
"name1": "Maier",
"name2": "Mayer",
"fuzzy_score": 80,
"phonetic_score": 100,
"combined_score": 88,
"metaphone_match": true,
"match": true,
"thresholds": {
"fuzzy_threshold": 75,
"metaphone_required": true
}
}

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Hi Jason,

this is what you´re doing as a warm up;)

Thanks a lot for your support, but in a first step I´ll stick to Genesys internal functionality. Even with test system on customer side for the provision of the data I want to be careful. One never knows how test data is created...

Hi Sophie -
I needed a warm-up this morning so I made a little test AWS lambda function that does this. 

If you promise not to use it for prod (laughing) I can give you the URL and config to make it work as well as a temporary token.

Here's the results:
{
"name1": "Maier",
"name2": "Mayer",
"fuzzy_score": 80,
"phonetic_score": 100,
"combined_score": 88,
"metaphone_match": true,
"match": true,
"thresholds": {
"fuzzy_threshold": 75,
"metaphone_required": true
}
}

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

If you wanted to go really old school, you could convert the values using Soundex rules (Soundex - Wikipedia).  You should be able to do all of the logic within Architect by looping through the letters of each string.  It is a little archaic, but I have used the logic in other (non-Genesys) applications in my past and it worked pretty well.

Hi Jason,

this is what you´re doing as a warm up;)

Thanks a lot for your support, but in a first step I´ll stick to Genesys internal functionality. Even with test system on customer side for the provision of the data I want to be careful. One never knows how test data is created...

Hi Sophie -
I needed a warm-up this morning so I made a little test AWS lambda function that does this. 

If you promise not to use it for prod (laughing) I can give you the URL and config to make it work as well as a temporary token.

Here's the results:
{
"name1": "Maier",
"name2": "Mayer",
"fuzzy_score": 80,
"phonetic_score": 100,
"combined_score": 88,
"metaphone_match": true,
"match": true,
"thresholds": {
"fuzzy_threshold": 75,
"metaphone_required": true
}
}

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Very interesting.  So, you have a bot or Virtual Agent listen for name, then do the conversion of characters and compare that with the value in the account.  I'll have to try this.

If you wanted to go really old school, you could convert the values using Soundex rules (Soundex - Wikipedia).  You should be able to do all of the logic within Architect by looping through the letters of each string.  It is a little archaic, but I have used the logic in other (non-Genesys) applications in my past and it worked pretty well.

Hi Jason,

this is what you´re doing as a warm up;)

Thanks a lot for your support, but in a first step I´ll stick to Genesys internal functionality. Even with test system on customer side for the provision of the data I want to be careful. One never knows how test data is created...

Hi Sophie -
I needed a warm-up this morning so I made a little test AWS lambda function that does this. 

If you promise not to use it for prod (laughing) I can give you the URL and config to make it work as well as a temporary token.

Here's the results:
{
"name1": "Maier",
"name2": "Mayer",
"fuzzy_score": 80,
"phonetic_score": 100,
"combined_score": 88,
"metaphone_match": true,
"match": true,
"thresholds": {
"fuzzy_threshold": 75,
"metaphone_required": true
}
}

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Yeah, you'll have to convert the output from the bot into a Soundex value and also convert the string from your external system into a Soundex value and then do your comparison of the two to see if you get a match.

Very interesting.  So, you have a bot or Virtual Agent listen for name, then do the conversion of characters and compare that with the value in the account.  I'll have to try this.

If you wanted to go really old school, you could convert the values using Soundex rules (Soundex - Wikipedia).  You should be able to do all of the logic within Architect by looping through the letters of each string.  It is a little archaic, but I have used the logic in other (non-Genesys) applications in my past and it worked pretty well.

Hi Jason,

this is what you´re doing as a warm up;)

Thanks a lot for your support, but in a first step I´ll stick to Genesys internal functionality. Even with test system on customer side for the provision of the data I want to be careful. One never knows how test data is created...

Hi Sophie -
I needed a warm-up this morning so I made a little test AWS lambda function that does this. 

If you promise not to use it for prod (laughing) I can give you the URL and config to make it work as well as a temporary token.

Here's the results:
{
"name1": "Maier",
"name2": "Mayer",
"fuzzy_score": 80,
"phonetic_score": 100,
"combined_score": 88,
"metaphone_match": true,
"match": true,
"thresholds": {
"fuzzy_threshold": 75,
"metaphone_required": true
}
}

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

I would assume like confidence, there could be a percentage of difference that is acceptable to match the account name.

Yeah, you'll have to convert the output from the bot into a Soundex value and also convert the string from your external system into a Soundex value and then do your comparison of the two to see if you get a match.

Very interesting.  So, you have a bot or Virtual Agent listen for name, then do the conversion of characters and compare that with the value in the account.  I'll have to try this.

If you wanted to go really old school, you could convert the values using Soundex rules (Soundex - Wikipedia).  You should be able to do all of the logic within Architect by looping through the letters of each string.  It is a little archaic, but I have used the logic in other (non-Genesys) applications in my past and it worked pretty well.

Hi Jason,

this is what you´re doing as a warm up;)

Thanks a lot for your support, but in a first step I´ll stick to Genesys internal functionality. Even with test system on customer side for the provision of the data I want to be careful. One never knows how test data is created...

Hi Sophie -
I needed a warm-up this morning so I made a little test AWS lambda function that does this. 

If you promise not to use it for prod (laughing) I can give you the URL and config to make it work as well as a temporary token.

Here's the results:
{
"name1": "Maier",
"name2": "Mayer",
"fuzzy_score": 80,
"phonetic_score": 100,
"combined_score": 88,
"metaphone_match": true,
"match": true,
"thresholds": {
"fuzzy_threshold": 75,
"metaphone_required": true
}
}

Hey Jason,

thanks for your response.

I know there is something like fuzzy matching I should use. However, right at the moment I just need to implement a PoC and I was hoping for some kind of creative idea, how I could do it with OOTB elements within Genesys. It is totally fine if this solution doesn´t cover each edge case and doesn´t reach the quality of a proper solution.

Regarding your question: There is also an authorization step based on account number. However, as the authorization has several steps incl. name and address, this shall be also part of the bot flow in order to take this effort from the agents.

Hey Sophie -
I don't think there's anything on platform to do that. Depending on the PII/GDPR/etc. needs you could always stand up a very cheap simple web service to do something like this for you. For example in Python3 you could write a Flask app to use metaphone and rapidfuzz to get two different comparisons and see if it's over your desired threshold. It's a neat problem. Out of curiosity is there not a better identifier you could use to verify the caller like account number, etc.? It sounds like you don't for your use case but I'm just curious.

.Jason

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

What I like to do is take the name and address from the outside system and use it as a Dynamic List and have the bot check against that.  If it matches in its "fuzzy" logic, then you just use the original information.  If not, you go with a Virtual Agent and have them spell it out.

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Hi Robert,

do you have any experience with dynamical lists in a comparable context? I started with a test set, but I came to the conclution that it is more strictly than a regular expression. It is also limited to 100 values, which is not really sufficient for name varieties, etc.

With the regular expression I can recognize names quite well, but still have the challenge with the check against salesforce.

What I like to do is take the name and address from the outside system and use it as a Dynamic List and have the bot check against that.  If it matches in its "fuzzy" logic, then you just use the original information.  If not, you go with a Virtual Agent and have them spell it out.

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Here is a link to dynamic slots:  Use custom dynamic list slot types in a bot flow - Genesys Cloud Resource Center

Yes, 100 is a limit, but that is the reason to use it.  Just get identification information and use that to create the collection for the lookup.  Then use Ask for Slot.  Then see if there is a match.  You can look at the confidence of the match and decide how close you want to the original for acceptance or not.  Make sense?

Hi Robert,

do you have any experience with dynamical lists in a comparable context? I started with a test set, but I came to the conclution that it is more strictly than a regular expression. It is also limited to 100 values, which is not really sufficient for name varieties, etc.

With the regular expression I can recognize names quite well, but still have the challenge with the check against salesforce.

What I like to do is take the name and address from the outside system and use it as a Dynamic List and have the bot check against that.  If it matches in its "fuzzy" logic, then you just use the original information.  If not, you go with a Virtual Agent and have them spell it out.

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

No, I am suggesting you use the dynamic slot as an on-the-fly dictionary of possible matches.  Look at the account and pull the name you want to match against into a collection and then use that for the dynamic slot.  When the caller says their name, it will check against the collection and provide a match or not.  You can take the confidence level of that match and decide if you feel it matches the original or not depending on that percentage.  

Spelling is better in a Virtual Agent as well as recognizing the name.  Takes $0.50 per call, but well worth it for recognizing names, emails, and alphanumeric input.

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Thank you for clarification, now I understand your approach. 

I tried to implement it and although I reduced confidence level from 10 (low) to 30 (high) the bot still struggles with recognition in case of hyphens. I replaced them, but still I see problems e.g. in case of different spelling. E.g. I tried as a street name "Alte Beroltzheimer Straße" and bot understands "alte beroldsheimer straße", which is quite similar, but still matching is failing. I guess upper cases make no difference? I also just tried out Beroltzheimer, but as bot understands "ds" instead of "tz", there is still no match.

Do you have an idea?

No, I am suggesting you use the dynamic slot as an on-the-fly dictionary of possible matches.  Look at the account and pull the name you want to match against into a collection and then use that for the dynamic slot.  When the caller says their name, it will check against the collection and provide a match or not.  You can take the confidence level of that match and decide if you feel it matches the original or not depending on that percentage.  

Spelling is better in a Virtual Agent as well as recognizing the name.  Takes $0.50 per call, but well worth it for recognizing names, emails, and alphanumeric input.

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Do you have the language for the initial flow set for de-DE?  Yes, the pz, pf, tz, ds is going to be problematic.  Did you test Virtual Agent with the recognition?  As for the failure route, you can still check the confidence of the last utterance and decide to keep it or not.

Thank you for clarification, now I understand your approach. 

I tried to implement it and although I reduced confidence level from 10 (low) to 30 (high) the bot still struggles with recognition in case of hyphens. I replaced them, but still I see problems e.g. in case of different spelling. E.g. I tried as a street name "Alte Beroltzheimer Straße" and bot understands "alte beroldsheimer straße", which is quite similar, but still matching is failing. I guess upper cases make no difference? I also just tried out Beroltzheimer, but as bot understands "ds" instead of "tz", there is still no match.

Do you have an idea?

No, I am suggesting you use the dynamic slot as an on-the-fly dictionary of possible matches.  Look at the account and pull the name you want to match against into a collection and then use that for the dynamic slot.  When the caller says their name, it will check against the collection and provide a match or not.  You can take the confidence level of that match and decide if you feel it matches the original or not depending on that percentage.  

Spelling is better in a Virtual Agent as well as recognizing the name.  Takes $0.50 per call, but well worth it for recognizing names, emails, and alphanumeric input.

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?

Do you have the language for the initial flow set for de-DE?  Yes, the pz, pf, tz, ds is going to be problematic.  Did you test Virtual Agent with the recognition?  As for the failure route, you can still check the confidence of the last utterance and decide to keep it or not.

Thank you for clarification, now I understand your approach. 

I tried to implement it and although I reduced confidence level from 10 (low) to 30 (high) the bot still struggles with recognition in case of hyphens. I replaced them, but still I see problems e.g. in case of different spelling. E.g. I tried as a street name "Alte Beroltzheimer Straße" and bot understands "alte beroldsheimer straße", which is quite similar, but still matching is failing. I guess upper cases make no difference? I also just tried out Beroltzheimer, but as bot understands "ds" instead of "tz", there is still no match.

Do you have an idea?

No, I am suggesting you use the dynamic slot as an on-the-fly dictionary of possible matches.  Look at the account and pull the name you want to match against into a collection and then use that for the dynamic slot.  When the caller says their name, it will check against the collection and provide a match or not.  You can take the confidence level of that match and decide if you feel it matches the original or not depending on that percentage.  

Spelling is better in a Virtual Agent as well as recognizing the name.  Takes $0.50 per call, but well worth it for recognizing names, emails, and alphanumeric input.

I need to compare personal data (name, address) captured via voice bot with the data provided from an external system. I want to implement that with OOTB possibilities of Genesys, but apparently there is not the possibility to compare both data based on phonetical similarity.

The problem is that I can use AreEqual() for the check, but this fails in case of different spelling of a name (Maier, Mayer, Meier, Meyer, etc.).

Of course I have the possibility to let the caller spell his/her name, but than I have the problem with spaces in case of several forenames, hyphens in case of double names and any kind of special characters.

Do you have an idea how I can solve that?