- From: Phil Ritchie <philr@vistatec.ie>
- Date: Fri, 14 Dec 2012 07:27:56 +0000
- To: public-multilingualweb-lt-comments@w3.org
- Message-ID: <OF760B2B6C.7ED0E689-ON80257AD4.0028E385-80257AD4.00290296@vistatec.ie>
I would like to propose the addition of "conformance" to Appendix C (Values for the Localization Quality Issue Type). The values in the appendix cover specific and discrete classes of error (putting "other" and "unintelligible" to one side). When you start to apply new text classification based quality checking methods to text several error classes may combine in subtle ways to produce a measure of quality that is "aggregate" across error types but none-the-less accurately indicative that something is wrong. For example, a target sentence may be deemed to have poor conformance when measured against a corpus of domain relevant reference translations. A score would reflect this poor conformance but the underlying errors within the sentence could be a mixture of grammar, spelling, style and/or terminology. In such instances you may not need to explicitly enumerate all of the combining errors and the extent of their contribution to the score, but just classify it under and umbrella term of "conformance". The proposed information for the "conformance" value would be as follows: Value conformance Description The content is deemed to have a level of conformance to a reference corpus. Reflects the degree to which the text conforms to a reference corpus given an algorithm which combines several classes of error type to produce an aggregate rating. Higher values reflect poorer conformance. Example "The harbour connected which to printer is busy or configared not properly." In a system which uses classification techniques this would be deemed to have poor conformance. The poor conformance is a function of the combined incorrect terminology, wrong spelling and bad grammar. Scope S or T Notes Reflects the degree to which the text conforms to a reference corpus given an algorithm which combines several classes of error type to produce an aggregate rating. Higher values reflect poorer conformance. Phil Ritchie ************************************************************ This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the sender immediately by e-mail. www.vistatec.com ************************************************************
Received on Friday, 14 December 2012 07:28:26 UTC