US2009055163A1PendingUtilityA1
Dynamic Mixed-Initiative Dialog Generation in Speech Recognition
Est. expiryAug 20, 2027(~1.1 yrs left)· nominal 20-yr term from priority
G10L 15/22G06F 40/56G10L 15/1822
43
PatentIndex Score
0
Cited by
0
References
0
Claims
Abstract
Disclosed are a method ( 500 ), apparatus ( 100 ) and computer program product for generating a mixed-initiative dialog to obtain information for dialog slots. A composite grammar dependent upon a set of unfilled slots is constructed ( 501 ). A prompt, dependent upon the a set of unfilled slots, is presented ( 309 ) to a user. An utterance is received ( 301 ) from the user in response to said prompt. Relevant information is determined based upon the further utterance. One or more said unfilled slots are filled ( 302 ) with said relevant information.
Claims
exact text as granted — not AI-modified1 . A method of generating a mixed-initiative dialog to obtain information for a set of information slots, the method comprising the steps of:
selecting a subset of slots from the set of information slots, dependent upon a set of unfilled slots for which information is to be obtained in a current dialog cycle; constructing a composite grammar dependent upon the selected subset of slots; generating a prompt dependent upon the selected subset of slots; receiving a user response to the prompt; comparing the response to the composite grammar; determining, dependent upon the comparing step, if the response provides relevant information for the set of unfilled slots; wherein said constructing step is dependent upon pre-defined atomic grammars each associated with at least one of the selected subset of slots and upon pre-defined grammar composition rules; said generated prompt forms; together with the composite grammar; and at least part of the current dialog cycle; and wherein the method comprises a further step of; filling one or more of the set of unfilled slots with said relevant information; and determining if execution flow of the current dialog cycle is complete; and if said flow is not complete, performing one of the steps of:
disambiguating the user response;
confirming the user response; and
repeating the selecting, constructing, generating, receiving, comparing an determining steps for a subsequent subset of slots,
wherein the selecting step is dependent upon at least one of a previous user response, a history of dialog-interaction which includes previous prompts and responses, a desired dialog flow, as yet unfilled slots in the set of information slots, and the selected subset of slots; wherein the pre-defined grammar composition rules specify; at least one of prefixes and suffixes which apply to a combination of the atomic grammars; and specific permutations of slots and associated atomic grammars; wherein a prefix is one of mandatory and optional; wherein a suffix is one of mandatory and optional; and wherein if certain permutations of slots are not valid or not desired, they can be specified as negative rules, wherein the constructing step is dependent upon at least one of:
a history of dialog-interaction which includes previous prompts and responses; and
a desired dialog flow,
wherein the generating step is dependent upon at least one of:
pre-defined slot information;
prompt generation rules; and
a desired dialog flow,
wherein the comparing step comprises matching said atomic grammars in the composite grammar against the user response, wherein the constructing of the composite grammar comprises the steps of:
specifying grammar composition rules;
selecting from the specified grammar composition rules those rules that apply to the selected subset of slots;
forming a slot grammar for each slot in the selected subset of slots using a corresponding rule in the selected grammar composition rules; and
combining the atomic grammars using the selected grammar composition rules to produce the composite grammar, said composite grammar being applicable to permutations of the selected subset of slots,
wherein the selecting of the subset of slots is dependent upon a state of a dialog with said user and pre-defined dialog flow information, wherein said generating step comprises the steps of:
defining rules for prompt generation;
using the rules, identifying a prompt segment associated with each slot in the selected subset of slots; and
combining the prompt segments, in an order specified by the rules, to generate the prompt for the selected subset of slots,
wherein said generating step is dependent upon at least one of pre-defined slot information, and dialog flow information; wherein said method further comprises acquiring slot information from a user input; wherein said method further comprises disambiguating acquired slot information; wherein said method further comprises confirming acquired slot information; wherein said method further comprises repeating the steps of: constructing the composite grammar; generating the prompt: receiving the user response; acquiring the slot information; disambiguating the acquired slot information; and confirming the acquired slot information, until all the unfilled slots in the set of information slots have been filled, wherein the user can specify the execution sequence of different phases of the dialog using a dialog flow strategy, wherein the dialog flow strategy comprises:
definition of a sequence of execution of different steps of user-interaction comprising information collection for slots, disambiguation and confirmation;
definition of slots that form part of each user interaction;
definition of a method to be used for the particular user-interaction; and
definition of a termination condition for dialog flow execution,
wherein the slots to be filled are determined based on:
a current state of the dialog slots that have been filled or their disambiguation state or their confirmation state;
the slot information; and
the dialog flow strategy.
2 - 26 . (canceled)Join the waitlist — get patent alerts
Track US2009055163A1 — get alerts on status changes and closely related new filings.
We store only your email — no account needed. See our privacy policy.