Replying to text messages via automated voice search techniques
First Claim
1. A method for automatically generating a text message from an arbitrary speech input, comprising steps for:
- receiving a statistical language model trained using one or more sets of real-world text messages as training input;
receiving a text message database constructed using a subset including some or all of the text messages in the one or more sets of real-world text messages received as the training input;
receiving an arbitrary user speech input;
evaluating the speech input relative to the probabilistic model and text message database to return a set of one or more probable speech recognition hypotheses corresponding to the arbitrary user speech input;
identifying a set of one or more text messages from the text message database as probabilistic matches to one or more of the speech recognition hypotheses;
ranking the probabilistically matching text messages in an order corresponding to a probabilistic confidence score associated with each speech recognition hypothesis;
selecting the highest ranked text message to paraphrase the arbitrary user speech input; and
transmitting the single selected text message to one or more recipients.
2 Assignments
0 Petitions
Accused Products
Abstract
An automated “Voice Search Message Service” provides a voice-based user interface for generating text messages from an arbitrary speech input. Specifically, the Voice Search Message Service provides a voice-search information retrieval process that evaluates user speech inputs to select one or more probabilistic matches from a database of pre-defined or user-defined text messages. These probabilistic matches are also optionally sorted in terms of relevancy. A single text message from the probabilistic matches is then selected and automatically transmitted to one or more intended recipients. Optionally, one or more of the probabilistic matches are presented to the user for confirmation or selection prior to transmission. Correction or recovery of speech recognition errors avoided since the probabilistic matches are intended to paraphrase the user speech input rather than exactly reproduce that speech, though exact matches are possible. Consequently, potential distractions to the user are significantly reduced relative to conventional speech recognition techniques.
-
Citations
20 Claims
-
1. A method for automatically generating a text message from an arbitrary speech input, comprising steps for:
-
receiving a statistical language model trained using one or more sets of real-world text messages as training input; receiving a text message database constructed using a subset including some or all of the text messages in the one or more sets of real-world text messages received as the training input; receiving an arbitrary user speech input; evaluating the speech input relative to the probabilistic model and text message database to return a set of one or more probable speech recognition hypotheses corresponding to the arbitrary user speech input; identifying a set of one or more text messages from the text message database as probabilistic matches to one or more of the speech recognition hypotheses; ranking the probabilistically matching text messages in an order corresponding to a probabilistic confidence score associated with each speech recognition hypothesis; selecting the highest ranked text message to paraphrase the arbitrary user speech input; and transmitting the single selected text message to one or more recipients. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
-
-
14. A system for selecting a text message for paraphrasing an arbitrary speech input, comprising:
-
a device for training one or more probabilistic language models from one or more sets of text messages; a device for constructing a text message database from the one or more sets of text messages; a device for receiving an arbitrary speech input; a device for generating one or more speech recognition hypotheses by applying the statistical language model to the arbitrary speech input; a device for identifying a group of one or more text messages in the text message database that probabilistically match each of the speech recognition hypotheses; a device for selecting a single message from the group of one or more text messages, said selected text message paraphrasing the arbitrary speech input; and a device for automatically transmitting the selected text message to one or more intended recipients. - View Dependent Claims (15, 16)
-
-
17. A computer-readable medium, which does not consist of a modulated data signal or carrier wave, having computer executable instructions stored therein for generating a text message based on an arbitrary speech input, said instructions comprising:
-
training a statistical language model and a corresponding text message database directly from one or more sets of real-world text messages, said language model including a lexicon limited to terms extracted from the one or more sets of real-world text messages; receiving an arbitrary speech input from a user; applying the statistical language model and lexicon to the arbitrary speech input to generate one or more probabilistic speech recognition hypotheses without requiring any user input to correct potential speech recognition errors; for each speech recognition hypothesis, identifying a group of one or more probabilistically matching text messages from the text message database, each probabilistically matching text message including a probability of match; ranking the probabilistically matching text messages based on the corresponding probability of match; evaluating the ranking of the text messages to select a single ranked text message to paraphrase the arbitrary speech input; and transmitting the selected text message to a recipient. - View Dependent Claims (18, 19, 20)
-
Specification