Dialogue systems Luděk Bártek			
Standards for Dialogue Interfaces Development			
		Dialogue systems	
W3C Voice Browser Activity SRGS SRGS XML Form ABNF form SISR		Luděk Bártek Laboratory of Searching and Dialogue, Fakulty of Informatics, Masaryk University, Brno spring 2023	
			
Standards for Dialogue Interfaces Development
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Objective - allow a dialogue interface transfer from a platform to another.
Standards for Dialogue Interface Development:
■ W3C Voice Browser Activity Standards,
■ AIML,
■ implementation platforms de-facto standards.
World Wide Web
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
1876 - The patent for the phone granted to A. G. Bell WWW
■ 1989 - The article " HyperText and CERN"(Tim Burnes Lee) circulates around the CERN.
■ Christmas 1990 - command line web browser and editor has been demonstrated.
■ 1991 - WWW general availability on computers in CERN
■ 1994 - The W3 Consortium first meting.
W3C Voice Browser Activity
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
1999 - the W3C Voice Browser Working Group formed.
Objective - proposal of standards allowing web browsing and access using a voice and phone.
Members:
■ HP
■ Nuance Communications
■ Lucent Technologies
■ Motorola
■ ScanSoft
■ IBM
■ Tel I me Networks
■ Vocalocity
W3C Voice Browser Activity Standards
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
VoiceXML - a dialogue strategy description language.
Speech Recognition Grammar Specification - the language for description of speech recognition grammars.
Semantic Interpretation for Speech Recognition - language for semantic interpretation support.
Speech Synthesis Mark-up Language - language for description of the sentence prosody factors for speech synthesis.
Standardy W3C Voice Browser Activity
Cont.
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Pronunciation Lexicon Specification - the pronunciation lexicon for a speech recognition and synthesis
Call Control XML - mark-up language used to control user - dialogue system connection.
State Chart XML - general use state machine description language.
Standardy W3C Voice Browser Activity
Processing
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
The standards are mark-up languages - must be interpreted.
There is a lot of implementation platforms:
■ Free available desktop platforms - JVoiceXML, PublicVoiceXML, ...
■ Commercial desktop - Optimtalk - there was a free available version for development and testing; LSD lab has a licence for computers in the lab.
■ Free available on-line platforms - Asterisk+VoiceGlue or OpenVXI, JVoiceXML...
■ Commercial on-line - Aspect Prophecy, Bevocal Cafe -can be used freely for development and testing purposes (2 parallels calls max)
Speech Recognition Grammar Specification
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
W3C specification of a language for description of context free grammars supporting the speech recognition.
Current version 1.0.
The JSGF replacement.
There are 2 two different notations:
■ XML
■ Augmented Backus-Naur Form (ABNF).
The only difference is the notation not the power of expression.
Support of the notations is platform depended.
■ Commonly used is the XML form.
Context Free Grammars
The Formal Languages Theory
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Grammar G — (A/, Z, P, S)
■ N - finite set of non-terminal symbols
■ Y. - finite set of terminal symbols (the language alphabet)
■ P - a set of rules
■ S - grammar root non-terminal symbol Context Free Grammar:
■ grammar G = (A/, Z, P, S)
■ where the rules are in the form N —>► {N U Z}*
SRGS grammar
XML form
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Starts with an XML prologue
■ <?xml version=" 1.0"encoding=". .." ?>.
Root element - grammar, contains the set of rules (rule elements).
Attributes
■ version - used SRGS version (current 1.0).
■ xmhlang- grammar language code.
■ root - rule id corresponding the the root non-terminal symbol.
■ mode - the communication mode of the grammar:
■ dtmf - using the DTMF codes
■ voice - voice grammar; the implicit value.
Rule Notation
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Element rule:
■ attributes:
■ id - the rule id (corresponds to the rule left-side non-terminal symbol).
■ Content - right side of the rule:
■ textual content - a sequence of terminal symbols.
■ element ruleref- non-terminal symbol; referenced by uri attribute.
■ element one-of- right side variants (operator |).
■ element item - logical division of the sequence; allows to specify the the count of repeating of particular part of the speech for example.
Sequence of Terminal and Non-terminal Symbols. SAMPLE^ I like KIND SRGS form.
<rule id=11 sample11 > I like <ruleref uri="#typ"/> SRGS form.
</rule>
Right side of the rule can be divided into the logical parts: <rule id=11 address11 > <item repeat="0-l">
<ruleref uri="#server"/> </item>
<item repeat="l-5">
dot <ruleref uri=,,#domainn/> </item> </rule>
Rule Variants
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Element one-of.
It allows to specify different variants of expected inputs.
The variants are enclosed in element item.
Example:
<rule id=11 colors11 > <one-of> <item>red</item> <item>green/item> <item>blue</item> </one-of> </rule>
Repeating
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
It allows to specify:
■ the optional parts of the speech
■ the recurrent parts of the speech
Notation - using the the attribute repeat of the element item.
Repeating count possible specification:
■ n times- n\
<item repeat="2">repeating</item>
■ < m. n > times- m-n
<item repeat="0-l">
I'd like to </item>
■ < a?, oo) times- m-
<item repeat="l-">Hi</item>
	Special Rule	
Dialogue	■	Can be used to input:
systems		■ the unspecified utterance - GARBAGE
Luděk Bártek		■ the unspeakable rule (prohibition of particular utterance) -
Standards for		VOID
Dialogue Interfaces		■ ever valid rule (even empty) - NULL
Development	■	They are used as special non-terminal symbols:
W3C Voice Browser		<ruleref spe c i al=" GARBAGE" / >
Activity		
SRGS XML Form ABNF form	■	Usage example:
SISR		<rule id=11 connection"> <ruleref special="GARBAGE"/> from <ruleref uri="#place"/> to <ruleref uri="#place"/> using <ruleref uri="#transport"/> </rule>
The SRGS ABNF Form
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Pure plain-text grammar form based on the BNF form.
<spojeni> : := I want to go from <from> to <whe:
at <when>11.11 <when> ::= <day> <time> <what> ::= train I bus
BNF based form is used by JSGF for example.
ABNF Form of SRGS
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Grammar header - may contain:
■ grammar natural language specification
■ grammar mode - voice/dtmf
■ root non-terminal
■ . ..
Grammar rules
■ form - Snon-terminal = (non-terminal\terminal)
■ non-terminal = XML form rule id.
*
SRGS Grammar ABNF form Header Structure
Dialogue systems
Luděk Bártek
Starts with document type specification. ■ #ABNF SRGS-version grammar.encoding #ABNF 1.0 ISO-8859-2
SRGS XML Form ABNF form
Followed by:
■ the root non-terminal symbol specification - root Snon-terminal;
■ grammar native language - language language code]
language en-US;
grammar usage mode - mode (voice|dtmf);
SRGS Grammar ABNF Notation
Dialogue systems
Luděk Bártek
Sequence - sequence of white-space separated terminal and non-terminal symbols:
$greating = Good day;
$date = $day th of $month $year;
SRGS XML Form ABNF form
Variants - corresponding sequences of terminal and non-terminal symbols separated by symbol
Stransport = bus I train;
Repeating:
■ optional parts of the speech - enclosed to '[ ]'
■ m—n - <m-n>
ABNF Form of SRGS Grammar Example
Dialogue systems
Luděk Bártek
#ABNF 1.0 UTF-8; root $url; language en-US; mode voice;
SRGS XML Form ABNF form
$url = [$protocol] [$server] dot
(Sdomain dot)<l-3>
$tld[$path]; Sprotocol = http I ftp I telnet I gopher $path = (/ Sdirectory ) <l-> /[$file];
Semantic Interpretation for Speech Recognition
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Semantic - assigns the meaning to the expressions/sentences.
Dialogue system and semantic:
■ assigns the interpretation to the utterances and to their parts
■ allows to extract relevant data from speech. SISR - W3C Voice Browser Activity Standard
■ serves for semantic interpretation of utterances
■ published on April 2007
■ current version 1.0.
■ is closely related to standards:
■ ECMA Script - the semantic interpretation uses the ECMAScript expressions
■ SRGS - the semantic evaluation is assigned to the SRGS rules using the rules attributes.
■ JSON - the interpretation is internally represented using a JSON objects.
Assignment of Interpretation to the Part of the Speech
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Semantic interpretation tends to be part of the SRGS rules.
Assigned to the rule using either the tag or the element tag:
■ SRGS XML Form: ■ element tag:
<item>
<ruleref uri=,,aggrement,,/> <tag>{out =,yes'}</tag> </item>
attribute tag:
<item tag=,,yes,,>Yeah</item>
SRGS ABNF Form:
■ interpretation follows the interpreted part of the speech.
■ from: {interpretation}
$claim = $aggre {yes} I $disagree {no}
Deriving the Interpretation of Whole Using Interpretations of Parts
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
Derivation notation - using ECMAScript expressions.
Assignment of derivation to the rules using either element of attribute tag.
Resulting interpretation is represented as JSON object.
Utterances evaluation:
■ partial interpretation access - rule right side symbols meaning access:
■ shadow variable rules attributes
■ the attribute N corresponds to the non-terminal N.
■ to return the interpretation into to the left-side
non-terminal - object out.
■ to return interpretation to the dialogue:
■ out object attributes
■ to the input field N corresponds attribute N.
	Utterances Evaluation			
	XM L form			
Dialogue systems	<rule id="ownership">			
Luděk Bártek	<item>			
Standards for Dialogue Interfaces Development	Mám <item repeat="0-l"> <ruleref uri="#color"/>			
W3C Voice Browser Activity SRGS SRGS XML Form	</item> <ruleref uri=,,#transport,, <tag>	>		
ABNF form SISR	{			
	out= rules.color+ ";" +	rules	.transport;	
	}			
	</tag>			
	</item>			
	</rule>			
	• • •	□	►       < i ► < i ►	
	Utterances Evaluation
	ABNF Form
Dialogue systems	
Ludek Bärtek	
Standards for Dialogue Interfaces Development W3C Voice Browser Activity SRGS SRGS XML Form	$ownership = I own   $color <0-l> $transport { out = rules.color+ ";" + rules.transport; >;
AtSNr Torm SISR	
Interpretation Assignment to the Input Fields
XML Form
Dialogue systems
Luděk Bártek
SRGS XML Form ABNF form
<rule id="ownership"> <item> I own
<item repeat="0-l"><ruleref uri="#color"/></item> <ruleref uri="#transpor"/> <tag> {
out.color = rules.color;
out.transport = rules.transport;
}
</tag> </item> </rule>
□ S1
	Interpretation Assignment to the Input Fields	
	ABNF form	
Dialogue systems		
Luděk Bártek		
Standards for Dialogue Interfaces Development	Sownership = I own $color <0-l> $transport	
W3C Voice Browser Activity	{ out.color = rules.color;	
SRGS SRGS XML Form ABNF form SISR	out.transport = rules.transport;	
		
	Literature	
Dialogue systems		
Luděk Bártek		
Standards for Dialogue Interfaces Development W3C Voice Browser Activity	■ SRGS Specification ■ SISR Specification	
SRGS SRGS XML Form ABNF form SISR	■ ECMAScript Specification ■ JSON Specification