Request for Proposal
Q: For Technical Scope 3a, please provide use-case scenarios for better understanding?
A: Our legal information server (server) will need to interact with a phone caller by recognizing (speech-to-text) spoken text in English, Vietnamese, and Spanish, and by speaking back to the caller (text-to-speech) in the same language. The server will expect to receive and generate text.
We anticipate that the phone system may include the capacity to do this itself, or may hand off the call as a SIP trunk to a third party such as Twilio, who would then interact with the caller via the trunk, and would interact with the legal information system via a REST API.
A possible scenario would be: (1) caller calls a phone number; (2) server is notified of the new call; (3) server “says” “Welcome to our system. What is your ZIP code?”; (4) server is sent the text of the speech; (5) server recognizes a ZIP code and proceeds with further questions, OR (6) server prompts again until it receives something it understands; (7) server “says” the next question or information to the caller.
Q: For Technical Scope 3b, do you already have a an existing server implementing Natural Language Understanding (NLU)? If so, what is it?
A: We are implementing Natural Language Understanding in a system called LegalServer, available from PS Technologies, Inc. (PSTI). In a proof of concept, PSTI has implemented “webhooks” to integrate with Twilio via its REST API, so Twilio can recognize speech from the caller and send text (speech-to-text) to the system, and can receive text (text-to-speech) from the system to speak back to the caller. The system will use its internal logic to understand what the caller is saying, and determine what next to say to or ask from the caller.