SSML Extensions for Chinese Voice Browsing

1
SSML Extensions for Chinese Voice Browsing

• Helen MENG, Wai-Kit LO, Tien-Ying FUNG, Yuk-Chi
  LI and Zhiyong WU Human-Computer Communications
  LaboratoryDepartment of Systems Engineering and
  Engineering ManagementThe Chinese University of
  Hong Kong
• 2nd November, 2005

2
Outline

• Characteristics of Chinese
• Proposed attributes for existing elements
• dialect-accent
• Proposed elements
• and
• Proposed attribute values
• for interpret-as attribute in element
• Summary

3
Characteristics of Chinese

• Rich in dialects, e.g., Cantonese, Shanghaiese,
  Mandarin
• Write alike, speak differently
• similar writing system e.g., ?? and ??
• significantly different pronunciations
• Mandarin with different accents
• No explicit phrase and word boundaries
• e.g., ??????????
• (we are) (now) (having) (a teleconference)
• proper segmentation is critical for prosodic
  control, pronunciation selection for homographs
  and resolution of semantic ambiguity
• Monosyllabic and tonal
• Syllable Lexical Tone ? lexical meaning of
  Chinese character
• tone can change according to meaning, context,
  mode of speaking

4
Phonetic Transcription Schemes

• Pronunciation of a character tonal syllable
• syllable tone
• Many transcription schemes developed for
  different dialects
• syllable in Roman alphabets
• tone as a one-digit Arabic number
• Popular schemes are
• pinyin (for Mandarin) ?? (bank) /yin2 hang2/
• jyutping (for Cantonese) ?? (bank) /ngan4 hong4/

5
Chinese Tone Systems
(2). ??/yang ping/,low levele.g., ?
(3). ?/shang/,risinge.g., ?
(4). ?/qu/,goinge.g., ?
(1). ??/yin ping/,high levele.g., ?

• Figure 1. Mandarin tone system (4 tones 1
  light tone)

(2). ??, high risinge.g., ?
(3). ??, high goinge.g.,?
(1). ??, high levele.g., ?
8(3). ??,middle enteringe.g., ?
9(6). ??,low enteringe.g., ?
7(1). ??,high enteringe.g., ?
(5). ??, low risinge.g., ?
(6). ??, low goinge.g., ?
(4). ??, low levele.g., ?
Figure 2. Cantonese tone system (9 tones,
specified in 6 classes)
6
dialect-accent
Beijing Mandarin
Guangdong Mandarin
Hong Kong Cantonese
7
Proposed dialect-accent Attribute

• Specify dialects and accents in a language
• use with xmllang XML1.0
• dialect-accent primary-subtag-optional-subtag
  
• primary-subtag 2ALPHA
• specify dialect
• e.g., MD for Mandarin, CT for Cantonese
• optional-subtag 2ALPHA
• specify accent
• e.g., BJ for Beijing, GD for Guangdong, HK for
  Hong Kong
• follows the abbreviations of Chinese provinces,
  autonomous regions and special administrative
  regions listed in the EDU.CN Domain Policy
  (??????????? EDU.CN ????????)1
• examples
• Mandarin in Beijing and Guangdong accent MD-BJ,
  MD-GD
• Cantonese in Hong Kong and Guangdong accent
  CT-HK, CT-GD

1 Defined by the China Education and Research
Network Information Centre (CERNET??????)
8
dialect-accent Attribute (continue)

• Hello, where are you from?
• ? (I am) ? (from) ?? (Beijing) ???
• ? (I am) ? (from) ?? (Guangdong) ???
• ? (I am) ? (from) ?? (Hong Kong) ???

Mandarin withBeijing accent
Mandarin with Guangdong accent
Cantonese with Hong Kong accent
9
and elements
10
Enrich , with ,

• Current SSML 1.0 and
• Proposed elements and
• Serve as cues for prosodic control (e.g., pause)
• Assist correct pronunciation selection for
  homographs
• A Cantonese example
• The character ? has FIVE pronunciations
• /haang4/ ?? (hiking)
• /hang6/ ?? (discipline)
• /hong2/ ?? (foreign trading company)
• /hong4/ ?? (bank)
• /hang4/ ?? (pedestrian)

11
Proposed Element

• Definition
• Defines the course of a Chinese phrase
• No attributes
• Occurs within
• These elements can be nested within
• , , , , ,
  , , , ,
• Example (an ancient poem) ????????
• Pessimistic phrasing
• ???? ????
• Optimistic phrasing
• ????? ???

Whole year unlucky
Not much money left
Only with a few unlucky events in the year
Have money left
12
Proposed Element

• Definition
• Defines the course of a Chinese word
• No attributes
• Occur within and
• These elements can be nested within
• , , , , ,
  , , ,
• Example ????????
• Segmentation 1
• ?? ??
• ?? ??
• Segmentation 2
• ??? ?
• ?? ??

/wui2/
1. This banquet is held as usual
This
banquet
as usual
hold
/wui3/
2. Tonight will be held as usual
Tonight
will
as usual
hold
13
element
14
Proposed Element

• Tone
• Important in Chinese pronunciation
• Tones can vary according to differences in
  meaning, context and mode of speaking
• ?
• in tone 2 means photo
• in tone 3 means facial appearance / minister
• Current SSML 1.0 phoneme
• Requires pronunciation transcription
• Example
• ph"soeng2"?
• ph"soeng3"?
• Proposed element
• with the required value attribute
• ? (photo)
• ? (face appearance)
• inherit the alphabet attribute, or explicitly
  specify

15
Examples of Using tone Element

• Tone changes on meaning
• ? (candy / sugar)
• ? (tone 2 /tong2/
  means candy)
• ? (tone 4 /tong4/
  means sugar)
• Tone changes on context
• ? (grandfather)
• ?? (tone 4 /je4/
  preceded by ?)
• ?? (tone 2 /je2/
  preceded by ?)
• Tone changes on mode of speaking
• ?? (English)
• ?? (tone 4 /man4/
  formal mode)
• ?? (tone 2 /man2/
  colloquial mode)

16
Values for interpret-as in
17
Proposed Legal Values for interpret-as Attribute

• VoiceXML2.0 Appendix P
• boolean, date, digits, currency, number, phone,
  time
• SSML 1.0 attribute values (W3C Working
  Group Note 2005)
• date, time, telephone, characters, cardinal,
  ordinal
• Propose 6 new values
• Chinese-name,
• fraction,
• measure,
• net,
• percentage,
• ratio

18
Chinese-name Value

• Specify as name to aid pronunciation selection
• ??? ? /daan1/ ? /sin6/ (surname)
• ?? /ming4 ming4/ ? /ming4 ming2/
  (given name)
• Format SG
• S surname, G given name
• Examples
• formatSG?? (Yao Ming)
• formatSGG??? (Sin Ming Ming)
• formatSSG??? (Au-yeung Sau)

19
fraction Value

• Specify as fraction
• e.g. 3/4
• Verbalization of fraction in Chinese
• with an additional word ?? (?out of)
• A / B (A out of B) B ?? A note that the order
  is reversed!
• e.g. 3/4 is verbalized as ?(four) ??(out of)
  ?(three)
• format and detail attributes not required
• Example
• ???3/4??
• (I) (ate) (orange)
• ???3/4??
  
• ????????? (I ate three-fourth of the orange)

20
measure Value

• Specify as measurement
• e.g. 10cm, 30ml
• measurement number unit
• number VoiceXML2.0 e.g. 10 is ten (not one
  zero)
• unit translated and pronounced in Chinese,
• e.g. cm is ??, g is ? , oz is ??, yd is ?
• format and detail attributes not required
• Example
• ?????180cm
• ?????180cms
• ??????????? (his height is 180cm)

(his) (height) (is)
21
net Value

• Specify as URI or email address
• Possible ways to verbalize a URI
• Read the whole string in English, including
  punctuations
• Omit http// (ftp//, etc.), read the rest in
  English
• Read alphabets in English, punctuations in
  Chinese
• format attribute value email or uri
• Example
• ????? http//www.w3.org
• (for details) (please) (browse)
• Possible verbalizations
• H T T P colon slash slash W W W dot W three dot O
  R G
• W W W dot W three dot O R G
• W W W ? W ? ? O R G (? dot ? three)
• Similarly the protocol part may be kept as
  another option
• ?????
  http//www.w3.org

22
percentage Value

• Specify as percentage
• Verbalization of percentage in Chinese
• with an additional word ??? (out of a hundred)
• A ???A
• e.g. 70 is verbalized as ???(out of a
  hundred)??(seventy)
• format and detail attributes not required
• Example
• ??????????70
• ??????????70say-as
• ??????????????? (ocean covers 70 of global
  surface)

(ocean) (covers) (global) (surface)
23
ratio Value

• Specify as ratio
• e.g. 13
• Verbalization of ratio in Chinese
• with an additional word ? (to)
• AB (A to B) A ? B
• e.g. 199 is verbalized as ?(one)?(to)???(ninety
  nine)
• format and detail attributes not required
• Example
• ?199 ??????
• ? 199
  ??????
• ???????????? (use diluted bleach at a ratio of
  199)

(use) (diluted) (bleach water)
24
Summary

• dialect-accent attribute to enrich the xmllang
  attribute
• and for text processing
• for pronunciation
• 6 values for interpret-as attribute
• Chinese-name
• fraction
• measure
• net
• percentage
• ratio

25
Thank You