Contents

Supported Say-as Text Normalization

In verbal conversations, certain categories of speech, such as currency and time, use a specific method to convey information. For example, when people read $12,345, it is usually spoken as "twelve-thousand-three-hundred-forty-five dollars" as opposed to "dollar-symbol one-two (pause) three-four-five", which is how a computer might interpret it.

In TTS, say-as text normalization directs the speech synthesizer to speak text in a specific manner so that the listener understands it. Without the say-as functionality, the synthesizer might speak a time of 10:30 AM as "one-zero-three-zero am". With say-as, the synthesizer can say the time as "ten-thirty a-m", which the listener understands easier.

The following table lists the say-as normalization types that are available and their support within Interaction Text to Speech.

Text normalization type	Usage	Supported	Notes
address	Processes mailing addresses	Yes	Processes US addresses, including military addresses, Post Office boxes, and rural routes. Abbreviations are expanded based on context: Abbreviation examples: 123 Main St. "St." spoken as "street" Springfield, IN "IN" spoken as "Indiana" PO Box "PO" spoken as "post office" RR 2 "RR" spoken as "rural route" Tips: Do not include the name of the addressee inside the `<say‑asinterpret‑as="address">` element. Instead, use regular text processing for all names. Use official United States Postal Service abbreviations for states, territories, and thoroughfare types. If you encounter the incorrect audio of "street" when "saint" is required, use the full word "Saint" instead of the "St." abbreviation, as seen in "St. Paul, MN". The say-as element for an address may contain newline characters Supported languages: en-US
alphanumeric	Spells letters and numbers	Yes	Supported languages:
alphanumeric	Spells letters and numbers	Yes	de-DE en-AU en-GB en-US es-US fr-CA	fr-FR it-IT ja-JP nl-NL pt-BR
boolean	Understands yes and no	Yes	This type uses the VoiceXML 2.0 defined type for Boolean. ITTS supports usage of `T`, `True`, `F`, and `False`, and local variants for this normalization type. Supported languages:
boolean	Understands yes and no	Yes	de-DE en-AU en-GB en-US es-US fr-CA	fr-FR it-IT ja-JP nl-NL pt-BR
currency	Processes currency amounts	Yes	In specifying currency values, you can use either the monetary symbol, such as `$`, or the associated abbreviation defined by ISO‑4217, such as (`USD`). ITTS processes values to the right of a decimal point for only four digits. ITTS ignores any additional digits. Supported languages:
currency	Processes currency amounts	Yes	de-DE (€ or EUR) en-AU ($ or AUD) en-GB (£ or GBP) en-US ($ or USD) es-ES (€ or EUR) es-US ($ or USD) fr-CA ($ or CAD) You can use a comma or a period as the decimal mark for fr-CA.	fr-FR (€ or EUR) it-IT (€ or EUR) ja-JP (¥ or JPY) nl-NL (€ or EUR) pt-BR (R$ or BRL) zh-CN (¥, CNY, or RMB)
date	Processes chronological dates	Yes	ITTS supports the following date formats: `mdy` - specify as `mm/dd/yyyy` in the text `dmy` - specify as `dd/mm/yyyy` in the text `ymd` - specify as `yyyy/mm/dd` in the text `md` - specify as `mm/dd` in the text `dm` - specify as `dd/mm` in the text `ym` - specify as `yyyy/mm` in the text `my` - specify as `mm/yyyy` in the text `y` - specify as `yyyy` in the text `m` - specify as `mm` in the text `d` - specify as `dd` in the text Example (en-US): `<say-as interpret-as="date" format="mdy">01/01/1984</say-as>` Output (en-US): "January first, nineteen eighty-four" Tip: You can use the following delimiters when specifying dates: `/` (slash) `-` (hyphen) `.` (period) You can also type single-digits for months and days, and two-digit years. Supported languages:
date	Processes chronological dates	Yes	de-DE en-AU en-GB en-US es-ES es-US fr-CA	fr-FR it-IT ja-JP nl-NL pt-BR zh-CN
digits	Reads strings digit-by-digit	Yes	Supported languages:
digits	Reads strings digit-by-digit	Yes	de-DE en-AU en-GB en-US es-US es-ES fr-CA	fr-FR it-IT ja-JP nl-NL pt-BR zh-CN
number	Reads strings as a value (not digit-by-digit)	Yes	Supported languages:
number	Reads strings as a value (not digit-by-digit)	Yes	de-DE en-AU en-GB en-US es-US es-ES fr-CA	fr-FR it-IT ja-JP nl-NL pt-BR zh-CN
ordinal	Processes ordinal numbers, such as "first", "second", and so on.	Yes	You can specify only the digits or the digits and an ordinal ending for the supported language. Note: If the language uses gendered forms for ordinals, and the gender is not specified in the text, ITTS hypothesizes the most-likely gender for the ordinal. Supported languages:
ordinal		Yes	de-DE en-AU en-GB en-US es-ES es-US it-IT ja-JP nl-NL pt-BR	fr-CA and fr-FR `1` results in "premier" or "première" with the ordinal. `1re` results in "première" `1 femme` results in "première femme" `1 cœur` results in "premier"
spell	Processes characters within a string, such as a word	Yes	The synthesizer speaks each character individually. Any punctuation characters in the string are read and named, such as ampersand (`&`), space ( ), and pound sign (`#`). Supported languages:
spell	Processes characters within a string, such as a word	Yes	de-DE en-AU en-GB en-US es-US es-ES fr-CA	fr-FR it-IT jp-JP nl-NL pt-BR zh-CN
telephone	Processes telephone numbers	Yes	The synthesizer reads telephone numbers, including SIP, according to local conventions in supported languages.
telephone	Processes telephone numbers	Yes	de-DE en-AU en-GB en-US es-US es-ES fr-CA	fr-FR it-IT ja-JP nl-NL pt-BR zh-CN
time	Processes time statements, such as "12:45 AM"	Yes	Supports hours, minutes, seconds, and 12 or 24-hour clock. This normalization type does not support any `format` options. This normalization type does not support durations, such as "60 minutes". For all languages, the separator for the hours and minutes of the time is a colon (`:`). For top-of-the-hour times, such as "2 o'clock", you can omit the colon and minutes. Example: `2 pm` Supported languages:
time	Processes time statements, such as "12:45 AM"	Yes	de-DE en-AU en-GA en-US es-ES es-US fr-CA/fr-FR These languages support usage of `h` as a separator between the hours and minutes, such as `1h30`.	it-IT ja-JP nl-NL pt-BR zh-CN