-4

I'm developing a skill on Amazon Alexa (Echo) for financial applications, but she isn't good enough at recognizing stock names. I require a maximum stock universe of about 15k stocks, and it obviously becomes very difficult for her to differentiate between the names of 15k non English dictionary strings. To solve this, I need a custom stack with:

  1. Custom grammar defined using backus naur form or similar (to improve my devices understanding of non English dictionary strings)
  2. Dynamic slots types (for example if we know the user is looking for a stock in their own portfolio, the options can be reduced from 15k to ~50)

Neither of these are offered by Amazon and I believe they are required for my project to progress. What options do I have other than the Amazon Echo?

Cheers

1 Answers1

1

We solved the same problem by sending the user speech text to our own NLP engine, where we extracted the Intent and Slot values.

We can easily extract the user speech text by creating new utterance like

userText {}

Here, userText is the intent. Whatever the user speak, the request will come as userText.

muTheTechie
  • 1,443
  • 17
  • 25