Capture block for Address from Alexa not working

I’m testing the PostalAddress slot and the StreetAddress slot in ADC, and it is not capturing the address as I was expecting

I’m using the capture block with PostalAddress slot and I also tried StreetAddress slot

and in ADC I am using the microphone

It seems it is capturing just as it is spoken e.g.
3,916 eagle run street
32 Marvin street
3,200 McKinney avenue

I was expecting the slot to capture true postal/street addresses like 3200 McKinney Avenue and not “3,200”

I am using an IF block after capture block to match the given address with that in our system but it is not matching because of the inconsistency with either the capture block or the postal/street address slots

Has anyone experience this or have suggestions as to capture the address correctly?

both of them are still beta.

Although I’m not English speaker, I tested with “2408 Leon Street” with pronounciation “twenty-four o eight leon street”, ended in disaster…

In my opinion and guess,

  • Alexa’s recognition, especially for proper nouns, seems not high even if using built-in slot types.
  • built-in slot type is a list of possible values maintained by Amazon. it’s like Amazon adds custom
    values. If users’ utterance is on the list, it works, but not on the list, does not work.
  • as workaround for it, you can add custom values for built-in slot type.
  • again, those are still beta, means not enough data for those.

Alexa will keep learning, so her recognition will improved in future. so I recommend these as the current workaround,

  • use choice block and create a built-in slot with adding custom values. capture block does not support adding custom values. (although custom values for built-in slot types in choice block seems not work right now…waiting for fixes…)

  • if you wanna get users’ address, you should use Device Address API with integration block. 2 advantages for this way:

    1. Generally, addresses has too much variations and Alexa cannot recognize those.
    2. Using its API, users does not need to speak srelatively-long sentences such as their address.
    3. Also, users don’t want to speak about their personal information.

Thanks for the suggestions and some of the options you mention will work for some use cases, especially if they are pre set options/info

I was going after a sort of virtual tour, for example visiting a college campus, stadium, or museum, etc you can input address and it would match to back end system address and info provided by the institution (amazon does have slots for these types of places but there are smaller/unique places it wouldn’t find)

I found the “FourDigitNumber” amazon slot works very well, it captures the actual digit. ended up going that route as it is more reliable than address - so now user can just say the unique four digit pin provided by the institution