Web triggered word set boosting for speech interfaces to the world wide web

Claim 3

A method according to claim 1 in which the computer system includes a local computer operable by the user, a remote computer at which the information is stored and a network link for coupling the local and remote computers, the displaying step including accessing the information from the local computer via a web browser.

Claim 4

A method according to claim 1 in which the information is stored in a hypertext markup language (HTML) source document and the step of forming a web triggered word set includes forming a set of words that includes a source word set including individual words extracted from the (HTML) source document.

Claim 5

A method according to claim 1 in which the step of forming a web triggered word set includes storing in a short term cache one or more previously-formed web triggered word sets for inclusion in a current web triggered word set.

Claim 6

A method according to claim 1 in which the step of forming a web triggered word set includes modifying the web triggered word set responsive to updating to display a new subset of the information.

Claim 7

A method according to claim 1 in which the step of forming a web triggered word set includes forming a set of words that includes words selected from at least one of a displayed subset of the information and a basic word set of command and function words chosen a priori.

Claim 8

A method according to claim 1 in which the step of performing a speech recognition process which includes the processing and modifying steps includes estimating a fit between an acoustic observation X of the speech input word sequence and a possible word sequence W according to a Bayes-type evaluation function F(S LM (W|H), S AC (X|W,H)), where S LM is a language model score, S Ac is an acoustic model score, and H is the web triggered word set.

Claim 9

A method according to claim 1 in which the step of performing a speech recognition process which includes the processing and modifying steps includes estimating a fit between an acoustic observation X of the speech input word sequence and a possible word sequence W according to an altered Bayes-like scoring function Pr(X|W)×Pr(W) Omega (W,H) using a special set of Omega(W,H) values for words W belonging to a predicted/web-triggered set H, so as to improve the scores of such words.

Claim 10

A computer system for user speech actuation of access to stored information, the system comprising: a computer including a central processing unit, a memory and a user input/output interface including a microphone for input of user speech utterances and audible sound signal processing circuitry; means for accessing and storing information in the memory of the computer system; a speech recognition processor operating on the computer system for recognizing words including individual words based on the input speech utterances of the user in accordance with a set of language/acoustic model and speech recognition search parameters which produce a probability score for at least the individual words; means for forming a web triggered word set from a selected subset of information in the document, the web triggered word set including a plurality of individual words contained in the displayed subset; means for modifying the language/acoustic model and speech recognition search parameters dynamically using the web triggered word set to boost the probability score of at least the individual words; and means responsive to the speech recognition processor for generating a word string based on the probability score as boosted by the modifying means for input to the accessing and storing means to initiate a change in the information accessed.

Claim 11

A computer system for user speech actuation of access to stored information, the system comprising: a computer including a central processing unit, a memory and a user input/output interface including a microphone for input of user speech utterances and audible sound signal processing circuitry; means for accessing and storing information in the memory of the computer system; a speech recognition processor operating on the computer system for recognizing words including individual words based on the input speech utterances of the user in accordance with a set of language/acoustic model and speech recognition search parameters which produce a probability score for at least the individual words; means for forming a web triggered word set from a selected subset of information in the document, the web triggered word set including a plurality of individual words contained in the displayed subset; means for modifying the language/acoustic model and speech recognition search parameters dynamically using the web triggered word set to boost the probability score of at least the individual words; and means responsive to the speech recognition processor for generating a word string based on the probability score as boosted by the modifying means for input to the accessing and storing means to initiate a change in the information accessed.

Claim 12

A system according to claim 9 including: means for displaying a first portion of the selected subset of the information via the computer on a display viewable by the user, so that the user can formulate speech utterances based on the displayed portion of the information; and means for updating the display to show a second portion of the selected subset of the information in accordance with the word string determined from the speech recognition search.

Claim 13

A system according to claim 9 in which the information stored in the memory comprises a first document and the accessing and storing means includes means for directing the system to access a different document responsive to the word string.

Claim 14

A method according to claim 9 in which the computer system includes a local computer operable by the user, a remote computer at which the information is stored and a network link for coupling the local and remote computers, the displaying means including means for accessing the information from the local computer via a web browser.