Browse Prior Art Database

Method and System for Efficient Identification and Correction of Pronunciation Errors in Bulk Processing

IP.com Disclosure Number: IPCOM000028874D
Original Publication Date: 2004-Jun-05
Included in the Prior Art Database: 2004-Jun-05
Document File: 2 page(s) / 25K

Publishing Venue

IBM

Abstract

Building a list of the unknown pronunciation words in a speech development IDE, or in a desktop dictation system is common practice. Selecting and editing each word in the list is also a current practice of correcting the unknown pronunciations. This article is directed to a method of streamlining the pronunciation correction process, using auditory parsing of such a list.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 2

Method and System for Efficient Identification and Correction of Pronunciation Errors in Bulk Processing

When constructing a speech application a common task is to produce pronunciations for words (or phrases) for which no known pronunciation exists. In such cases a typical procedure entails selecting a single word (or phrase) whose pronunciation you wish to verify, playing the default pronunciation and then optionally working with a pronunciation editor to modify the pronunciation. The drawback to this approach is that it wastes a great deal of development time when interacting with a large set of pronunciations.

The present disclosure is a pronunciation validation and correction system whereby pronunciations are played through a speaker sequentially with a few seconds delay (or a pleasant tone) to clearly denote the breaks in words or phrases. The developer would only need to involve themselves with the system if they heard an incorrect pronunciation. By presenting the verification process in this way the developer is removed from tasks of selecting and verifying individual pronunciations and only has to focus on listening for incorrect pronunciations, which occur relatively infrequently.

The correction process of the unknown pronunciation list works a loop, with one iteration of the loop representing one word or phrase whose pronunciation is to be verified in a sequence:

1

Page 2 of 2

End of the list?

No

   Does pronunciation exit?

Yes

User stopped the process?

Y...