Conventions for Language Implementations

Phrase tags

Standard phrases should be labeled using the following conventions:

• The colon (:) can be used as the first character of a phrase to indicate that a phrase is a standard phrase. Custom phrases may not start with a colon.
• The letters of the alphabet, digits, the minus sign ("-"), and the touchtone symbols ( * )and ( # ), will normally be allocated fixed phrase numbers and have phrase numbers starting with a colon.
• Additional numbers and phrase fragments, such as the teens, the tens through 90, hundred, and thousand, as necessary to speak numbers in a natural style for a particular language are also denoted with a colon, and assigned fixed phrase numbers. Some languages may have to augment these with different phrases to allow speaking a number with the appropriate gender, or to include additional phrase fragments such as the numbers through 99.
• If a language implementation supports multiple inflections, then the standard phrases would be recorded with a rising and falling inflection, in addition to the medial inflection. The inflection is normally denoted by adding a period ( . ) for falling inflection or a question mark ( ? ) for rising inflection to the end of the phrase tag. For example, for US English, the phrase tag ":100?" refers to the fragment "hundred" spoken with a rising inflection:
• Other standard phrases necessary to speak money in the correct currency for a language, to speak dates and times, etc., must be defined. These can start with a colon and can be available in multiple inflections, but this is not required.
• All standard phrases must be included in the proto.pl file, whether or not they start with a colon.

Format names

It is desirable that an application be easily modified to support different languages. This requires that as many of the formats as possible follow a standard naming convention, so the formats will be available in all languages. The current naming convention is based on the US English implementation. Thus, it is desirable that each language support the same formats as US English using the same format name. Examples are THMAM (time: hour, minute, AM/PM), and DMSPDYY (date: with month spoken as a word -- for example, January -- day, and year as four digits). Currency is denoted with the dollar sign "$" (such as N$D2 -- number spoken as dollars and 2-digit cents).

In particular, the most standard way of speaking numbers, characters, dates, currency, and times should be assigned the default format for the language (N, C, D, N$, and T, respectively.)

Additional formats and names for formats can be specified to support natural spoken data according to each language's grammar. For example, the Spanish formats for numbers referring to a masculine, feminine, or neutral noun are "NM," "NF," and "NN," respectively.

If TTS is provided for a language, the TTS formats should be the same as the formats for prerecorded phrases, except that the format is preceded by the letter A (for example, "AN" and "ATHMAM"). Thus non-TTS formats should not start with the letter A.

Phrase numbers

The proto.pl file is used to define phrase numbers for the standard phrases. The phrase numbers 1-999 are reserved for standard phrases.

It is desirable, but not necessary that all standard phrases be allocated specific phrase numbers within this range. The phrases that must be allocated in this range are those that must be allocated a specific number or allocated contiguously for the implementation. Examples are the letters of the alphabet, numbers, months, and ordinal numbers if needed. The implementation will typically compute the phrase numbers for these by adding an offset to a base phrase number of a contiguous set of phrases for the set.

In cases where standard phrases are referenced only by their phrase tag, such a "dollars and" for US English, it is not necessary that a fixed phrase number be assigned. These can be assigned phrase number 0, in the proto.pl file.

Rules for creating script subroutine source files

A language package should include a set of script subroutine source (.t) files, necessary to implement the various formats, as well as the specific files listed previously. The rules for these files are based on the rules for external functions with the following differences:

• The files are installed in the same language directory as the language definition files, /vs/bin/ag/formats/language, rather than in the directory for external functions.
• The file names should start with an underscore ("_"), and use a .t suffix.
• The file names should include a language identifier, such as "CS" for Castilian Spanish ( _CSsp_dol.t).
• Labels within the file should be of the following format:

  L_ fileprefix _ uniquesuffix

  For example, labels within the above file could include L__CSsp_dol, L__CSsp_dolErr, L__CSsp_dol2, L__CSsp_dol_rtn, etc.

© 2006 Avaya Inc. All Rights Reserved.