- From: Jukka K. Korpela <jkorpela@cs.tut.fi>
- Date: Mon, 2 Apr 2007 08:50:46 +0300 (EEST)
- To: www-html@w3.org
On Mon, 2 Apr 2007, Nicholas Shanks wrote: > Two comments [on the read="..." attribute] > > I use <ssml:phoneme> elements in personal documents, which are manipulated > with XSLT before being sent to a speech engine, one option is to use > something compatible with that. People have expressed their concern about the interpretation of the attribute value. Defining it as IPA notation would be theoretically promising, but for most cases, that would be overkill. Usually it is sufficient to specify the pronunciation the same was as the language of the content of the element is written. So for example, if we wanted to specify that "I" is to be read as a Roman numeral in some context, we could write <span read="the first">I</span> in an English document and <span read="der erste">I</span> in a German document. This, by the way, might help automatic translation as well: it could know or guess that when translating from English, such an "I" is to be kept as is and not interpreted as a personal pronoun. The vast majority of authors doesn't know IPA, or knows it passively only (can follow pronunciation instructions in IPA notations but not write them). Besides, it wouldn't be a bad idea for a graphic browser to give users an optional access to read="..." attributes, e.g. the way Firefox lets you right-click on anything and select Properties, to see the language of the element (as declared in markup), its advisory title if present, its destination if it's a link, etc. In such usage, plain language is more useful to most people than IPA. As a policy issue, according to what I have understood from Unicode list discussions, many experts think that IPA should not have a special status among phonetic writing systems. There are other systems in use, even though IPA is the most common in linguistics. Thus, the pronunciation should be specified the same way as you would do in normal text. For example, if you would like to specify the pronunciation of some foreign word, you would try to write it according to the rules of the document's language. This indicates the pronunciation very coarsely, but often in a useful way. If an _additional_ attribute is defined for the purpose of giving pronunciation instructions, it might use IPA by definition, or it might be defined as using _some_ phonetic notation, to be defined separately (though this admittedly seems to result in some attribute spaghetti). > Also "read" has the problem of not knowing what tense it is (could be > homophonous with reed or red). May I suggest "pronounce" as an alternative > attribute name? Generally, verbs should be avoided in attribute names, since in a logical markup language, attributes are supposed to indicate properties or relationships, not actions. We have the precedent align="...", but it has been condemned to deprecation. In HTML 4.01 there's also accept="...", defer="...", and others. In this case, the word "read" would really reflect the _meaning_ of the attribute: how the content is read, or is to be read. Admittedly people could read it two ways when pronouncing it, but I don't think that's a serious problem. Most people would read it as if it were an imperative, which doesn't sound like descriptibe markup, but so what? And pronouncing attribute names isn't the main use of those names, and it's not part of the attributes meaning. When writing about them in HTML, you could always express your preference by using <code read="reed">read</code> or <code read="red">read</code>. :-) -- Jukka "Yucca" Korpela, http://www.cs.tut.fi/~jkorpela/
Received on Monday, 2 April 2007 05:50:56 UTC