|
Title:
|
IMPROVING AUTOMATIC SPEECH TRANSCRIPTION FOR MULTIMEDIA CONTENT |
|
Author(s):
|
Masood Masoodian , Bill Rogers , Saturnino Luz |
|
ISBN:
|
978-972-8924-44-7 |
|
Editors:
|
Pedro Isaías , Miguel Baptista Nunes and João Barroso (associate editors Luís Rodrigues and Patrícia Barbosa) |
|
Year:
|
2007 |
|
Edition:
|
V I, 2 |
|
Keywords:
|
Multimedia retrieval, multimedia indexing, speech recognition, computer-assisted transcription, transcript error
correction, usability. |
|
Type:
|
Full Paper |
|
First Page:
|
145 |
|
Last Page:
|
152 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Automatic Speech Recognition systems are increasingly being used in multimedia retrieval applications, where speech
recognition is used to aid the creation of high-quality transcripts for data such as multimedia meeting recordings, lectures
and presentations, video and audio libraries, and broadcast news. These transcripts are then used for indexing and
retrieval of the multimedia content over the Internet. Automatic transcription in these domains is however a challenging
task for a number of reasons, including unfavourable recording conditions, high frequency of out-of-vocabulary words
and multiplicity of speakers and accents. Furthermore, with the increasing volume of multimedia data, particularly video
and speech, being recorded, stored, and shared over the Internet, full manual correction of transcribed speech is
impractical. Hybrid transcription systems are therefore needed to allow combining automatic transcriptions for the most
part, and manual corrections to some extent, in generating accurate transcription of multimedia content. This paper
presents a graphical system which limits human involvement to correcting some, but not all, transcription errors. These
corrections can then be used to dynamically update the system vocabulary, thus helping the system to remove related
transcription errors automatically. |
|
|
|
|
|
|
