Apparatus for converting speech to text
First Claim
Patent Images
1. A transcription system for generating text for users in accordance with recorded speeches making use of automatic speech recognition, the system comprising:
- a plurality of user terminals for recording speeches;
at least one automatic speech recognition processor to generate text from recorded speeches; and
communication means operative to return corresponding text to respective users;
wherein at least one automatic speech recognition processor is adapted to improve recognition accuracy for a user using selected data of the recorded speeches and the corresponding texts of at least one other user;
said speeches each having an identifier associated with it corresponding to subject matter area and/or user accent; and
said data being selected and pooled dependent upon identifiers common to said speeches;
said automatic speech recognition processor having a different language model for each identifier; and
each language model being adapted using pooled data with the corresponding identifier.
3 Assignments
0 Petitions
Accused Products
Abstract
A speech-to-text conversion system which comprises a plurality of user terminals for recording speeches, at least one automatic speech recognition processor to generate text from recorded speeches, and communication links to return corresponding texts to respective users. At least one automatic speech recognition processor improves recognition accuracy using data of the recorded speeches and the corresponding texts, the data being selected dependent upon subject matter area.
326 Citations
22 Claims
-
1. A transcription system for generating text for users in accordance with recorded speeches making use of automatic speech recognition, the system comprising:
-
a plurality of user terminals for recording speeches;
at least one automatic speech recognition processor to generate text from recorded speeches; and
communication means operative to return corresponding text to respective users;
wherein at least one automatic speech recognition processor is adapted to improve recognition accuracy for a user using selected data of the recorded speeches and the corresponding texts of at least one other user;
said speeches each having an identifier associated with it corresponding to subject matter area and/or user accent; and
said data being selected and pooled dependent upon identifiers common to said speeches;
said automatic speech recognition processor having a different language model for each identifier; and
each language model being adapted using pooled data with the corresponding identifier. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A transcription method for generating text for users in accordance with recorded speeches making use of automatic speech recognition, comprising:
-
recording speeches at a plurality of user terminals;
using at least one automatic speech recognition processor to generate text from said recorded speeches; and
returning corresponding text to respective users;
whereinselected data of said recorded speeches and corresponding text of multiple users are used to adapt operation of at least one automatic speech recognition processor to improve recognition accuracy;
each speech being marked with an identifier corresponding to subject matter area and/or user accent;
said data being selected and pooled dependent upon identifiers common to said speeches;
a different language model being used by the automatic speech recognition processor for speeches having different identifier; and
each language model being adapted using pooled data with a corresponding identifiers.
-
Specification