Voice-enabled dialog system
View Patent ↗A voice-enabled help desk service is disclosed. The service comprises an automatic speech recognition module for recognizing speech from a user, a spoken language understanding module for understanding the output from the automatic speech recognition module, a dialog management module for generating a response to speech from the user, a natural voices text-to-speech synthesis module for synthesizing speech to generate the response to the user, and a frequently asked questions module. The frequently asked questions module handles frequently asked questions from the user by changing voices and providing predetermined prompts to answer the frequently asked question.
1. A voice-enabled spoken dialog service comprising:
an automatic speech recognition module comprising a general-purpose acoustic model and a domain-specific model, wherein the general-purpose model is used for bootstrapping at initial deployment of the spoken dialog service and the domain-specific model is used to adapt the automatic speech recognition module after deployment, wherein a language corpus for the domain-specific acoustic model is drawn from at least one domain-specific website;
a spoken language understanding module that performs text normalization, entity extraction and semantic classification using a boosting approach that balances human-crafted rules with available data;
a dialog management module that comprises an interpreter, finite state machine engine and an action template;
a text-to-speech synthesis module for synthesizing speech; and
a frequently asked questions module, wherein the spoken dialog service communicates with the user in a first voice, and wherein the frequently asked questions module performs:
receiving an utterance from the user;
determining whether the utterance is a frequently asked question;
if the utterance is a frequently asked question, answering the frequently asked question in a second voice; and
after answering the frequently asked question in the second voice, providing a prompt in the second voice to return the user to the first voice.
2. The voice-enabled spoken dialog service of claim 1 , wherein the frequently asked questions module uses a question and answer table having data extracted from previous callers to a spoken dialog system.
3. The voice-enabled spoken dialog service of claim 2 , wherein the question and answer table is further partitioned into small subsets, each having a corresponding semantic tag, wherein if the user asks a question that closely matches a question found in the questions and answer table, the answer is automatically passed to the dialog management module with any entities and semantic tags.
4. The voice-enabled spoken dialog service of claim 1 , wherein the text-to-speech synthesizer module synthesizes speech using a screenwriting dialog technique where a back story is created for the synthesized voice based on a desired character trait.
5. The voice-enabled spoken dialog service of claim 4 , wherein the desired character trait is cheerful.
6. The voice-enabled spoken dialog service of claim 4 , wherein the desired character trait is trustworthy.
7. The voice-enabled spoken dialog service of claim 4 , wherein the desired character trait is helpful.
8. The voice-enabled spoken dialog service of claim 4 , wherein the desired character trait is calm.
9. The voice-enabled spoken dialog service of claim 4 , wherein the desired character trait is strong-willed.
10. The voice-enabled spoken dialog service of claim 1 , wherein the second voice is of a different gender from the first voice.
11. The voice-enabled spoken dialog service of claim 1 , wherein the second voice has different characteristics from the first voice.
12. The voice-enabled spoken dialog service of claim 1 , wherein if the frequently asked question is a first frequently asked question asked by the user, the method further comprises:
presenting a first frequently asked question prompt in the second voice.
13. The voice-enabled spoken dialog service of claim 1 , wherein the first frequently asked question prompt instructs the user that a virtual agent associated with the second voice handles the frequently asked questions.
14. The voice-enabled spoken dialog service of claim 12 , wherein the frequently asked questions module further performs:
after answering the frequently asked question in the second voice, providing a prompt in the second voice to return the user to the first voice.
15. The voice-enabled spoken dialog service of claim 1 , wherein the frequently asked questions module further performs:
if the utterance is a frequently asked question, determining whether the frequently asked question is a first frequently asked question by the user;
if the utterance is the first frequently asked question asked by the user:
providing a first frequently asked question introduction prompt using the second voice; and
answering the frequently asked question in the second voice; and
if the utterance is not the first frequently asked question asked by the user:
providing an introductory prompt different from the first frequently asked question introductory prompt in the second voice; and
answering the frequently asked question in the second voice.
16. The voice-enabled spoken dialog service of claim 15 , wherein the frequently asked questions module further performs:
providing a prompt in the first voice that a first voice agent associated with the first voice has returned to continue interacting with the user.