Home > Lucent Technologies > Communications System > Lucent Technologies INTUITY CONVERSANT System Version 6.0 Guide

Lucent Technologies INTUITY CONVERSANT System Version 6.0 Guide

    Download as PDF Print this page Share this page

    Have a look at the manual Lucent Technologies INTUITY CONVERSANT System Version 6.0 Guide online for free. It’s possible to download the document as PDF or print. UserManuals.tech offer 413 Lucent Technologies manuals and user’s guides for free. Share the user manual or guide on Facebook, Twitter or Google+.

    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Introduction to Voice Response Application Design 
    Page 1-5 Important Terminology 
    1
    Important Terminology
    It is a good  id ea to b ec ome familiar with the some of the terms and  c onc ep ts 
    p resented in the rest of this b ook.  The list below is p rovid ed  to help  fac ilitate your 
    familiarization.
    announc ement Sp eec h played  b y the system to the c aller that informs, b ut 
    d oes not instruc t the c aller to ac t.  Comp are to 
    prompt.
    ap plic ation The c omp uter p rog ram that d efines and  c ontrols the voic e 
    response transac tion b etween the system and  the c aller.
    barge-In The capability that allows callers to respond while a prompt 
    is b eing  p layed .  This is similar to d ial throug h for touc h 
    tones.
    c aller The p erson who c alls for a servic e, g ets c onnec ted  to the 
    system, and interac ts with an app lic ation.
    c onnec ted -d ig it A sequenc e of d ig its sp oken b y a c aller without intentional 
    or reg ular p auses in b etween d ig its.
    d ial ahead The touc h-tone rec og nition c ap ab ility that allows the 
    system to c ollec t touc h tones as they are entered  b y 
    c allers, even b efore they are asked  for.  The touc h-tone 
    input is then used in the order in whic h they were rec eived.  
    This allows c allers to resp ond  to more than one p romp t at a 
    time, without having to listen to the intermed iate p romp ts.
    d ial throug h The touc h-tone c ap ab ility that allows c allers to resp ond  
    while a prompt is being played.  The playback of speech 
    c eases and  the ap p lic ation resp ond s to the key that was 
    p ressed .  This is similar to b arg e-in for WholeWord  sp eec h 
    rec ognition.  Also known as 
    talk off.
    fax The c apab ility that allows an ap p lic ation to send  stored  or 
    d ynamically c reated  fax messag es at the caller’s request, 
    or rec eive inc oming  faxes from the c aller.
    Flex Word
     speech 
    rec og nitionThe op tional system c ap ab ility that rec og nizes c allers 
    spoken input based on matc hing caller speech to word 
    models fashioned from representations of sub-words 
    (phonemes), the smallest unit of sp eec h.
    Flex Word
     Toolkit An op tional software p ac kag e that p rovid es a 
    p oint-and -c lic k, grap hic al mec hanism for you to d efine 
    c ustom list of word s for use with FlexWord  Sp eec h 
    Rec og nition. 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Introduction to Voice Response Application Design 
    Page 1-6 Important Terminology 
    1
    g rammar The set of rules b y whic h speec h is rec og nized  b y 
    WholeWord  sp eec h rec og nition.  For examp le, a Promp t & 
    Collec t statement using  the US_1_5 g rammar will 
    rec ognize the word s “ one,”  “ two,”  “ three,”  “ four,”  and  “ five”  
    in US Eng lish.
    Grap hic al Sp eec h 
    Ed i t o rAn op tional software p ac kag e p rovid es a point-and -c lic k, 
    g rap hic al mec hanism for you to rec ord  and  ed it p romp ts 
    and  announc ements for use in your ap p lic ations.
    key word One of the word s in a list of word s that the system is 
    instruc ted  to rec ognize at a p artic ular p oint in the 
    transac tion.
    menu A p romp t that gives c allers a c hoic e of two or more op tions.  
    For examp le, “ For sales, p ress or say 1.  For servic e, p ress 
    or say 2.  For an attend ant, p ress or say 0.”
    p hrase sc reening The speec h rec ognition c ap ab ility that d ec id es whether or 
    not a c and idate key word  is a c lose enough matc h to b e 
    d ec lared  a valid  key word .
    Promp t & Collec t 
    actionThe ac tion used  to p lay a p romp t to the c aller, ac c ep t the 
    c aller’s resp onse, and  g o to the ap p rop riate p lac e in the 
    ap p lic ation to proc ess the c aller’s req uest.
    p romp t Sp eec h p layed  b y the system that instruc ts the c aller to 
    enter information that is p art of a Promp t & Collec t ac tion.  
    Comp are to 
    Announc ement.
    rec og nition The p roc ess within the system that c omp ares c aller sp eec h 
    to internal mod els and returns a matc h to the ap p lic ation.
    rec og nition typ e The c hoic es that are assoc iated  with the Rec og nition Typ e 
    field  in the Promp t & Collec t ac tion form found  in the Sc rip t 
    Builder program.  The values for recognition type, minimum 
    number of digits, and maximum of digits work in 
    c onjunc tion to allow the system to selec t a g rammar to b e 
    used  for that rec og nition event.  See also 
    grammar.
    Sc rip t Build er An optional software p ac kag e that allows you to d efine and  
    g enerate voic e resp onse ap p lic ations to run on the I
    NTUITY 
    C O N VERSA N T s y st e m .
    sp eec h, c ustom The p art of the system sp eec h datab ase that inc ludes 
    ap p lic ation-dep end ent, p rerec orded  sp eec h p hrases.
    sp eec h, enhanc ed 
    basicThe part of the system speech database that inc ludes 
    prerecorded speech phrases corresponding to numbers, 
    ord inal numb ers, d ays of the week, and  months of the year.  
    Ap p lic ations use these p hrases to sp eak numb ers, 
    amounts, dates, and  times in a natural-sound ing  way.
    sp eec h, p rerec ord ed A promp t or announc ement that has been rec ord ed  b y a 
    person. 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Introduction to Voice Response Application Design 
    Page 1-7 Important Terminology 
    1
    sub stitution error An error mad e b y the FlexWord  speec h rec og nition 
    software where it mistakes one word  for another.
    talk off See 
    dial throug h.
    Text-to-Speec h An op tional software p ac kag e that c onverts ASCII text into 
    sp oken, c omp uter-g enerated  p romp ts and 
    announc ements.  This pac kag e is supp orted  for the US 
    Eng lish lang uag e only.
    touc h-tone The sig nal sent when a c aller p resses any of the 12 keys on 
    a p ush-b utton telep hone that send s d ual tones rather than 
    rotary p ulses.
    transac tion  The exc hang e of information b etween the c aller and  an 
    ap p lic ation.  In a typic al transac tion, the c aller d ials in to or 
    g ets transferred  to the system, then the system answers 
    and  plays a g reeting .  The c aller enters information in 
    response to sp oken p romp ts and  the system sp eaks 
    information b ac k until the interac tion is c omp lete.  See also 
    application.
    usability The system q uality that reflec ts whether or not c allers c an 
    learn and  use the features suc c essfully.  A system with 
    hig h usab ility is c alled  
    easy to use, or usable.
    user Interfac e The asp ec t of the app lic ation with whic h c allers interac t; 
    inc lud es p romp ts and announc ements from the system to 
    whic h c allers resp ond  using  touc h-tone or sp eec h.
    voc abulary The set of wordlists assoc iated with a partic ular FlexWord 
    application.
    WholeWord  speec h 
    rec og nitionThe op tional system c ap ab ility that rec og nizes the 
    lang uag e-sp ec ific  word s (and  c ommon synonyms) for the 
    d ig its 0 throug h 9.  Rec og nition is b ased  on matc hing  
    c aller sp eec h to word  mod els fashioned  from many 
    samp les of p eop le saying  eac h entire, 
    whole word .
    word A FlexWord  sp eec h rec og nition voc ab ulary item c onsisting  
    of either a sing le word  or a p hrase of several word s.
    word spotting The speech recognition capability that allows the system to 
    p ic k out key word s from a stream of c aller sp eec h, whic h 
    may inc lud e extraneous sp eec h, b ac kg round noise, or 
    c aller noises.  Works in c onjunc tion with p hrase sc reening .
    wordlist A set of words used  with FlexWord  sp eec h rec og nition that 
    can be recognized by the system. 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Introduction to Voice Response Application Design 
    Page 1-8 Important Terminology 
    1 
    						
    							Voice Response Advanced Technologies 
    Page 2-1 Overview 
    2
    INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    2
    2Voice Response Advanced 
    Technologies
    Overview
    This c hap ter p rovid es an introd uc tion to the INTUITY™ C ON VERSAN T®  s y st e m  
    ad vanc ed  tec hnolog ies inc lud ing  touc h-tone and  d ial p ulse rec og nition, sp eec h 
    rec ognition, Text-to-Sp eec h, and  the Sc rip t Build er FAX Ac tions.
    Purpose
    The purp ose of this c hap ter is to d esc rib e what eac h ad vanc ed  tec hnolog y d oes, 
    the typ es of ap p lic ations for whic h eac h is b est suited, and  how the tec hnolog ies 
    c an work tog ether. 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Voice Response Advanced Technologies 
    Page 2-2 Touch-Tone and Dial Pulse Recognition 
    2
    Touch-Tone and Dial Pulse 
    Recognition
    The INTUITY CONVERSANT system p rovid es touc h-tone rec og nition and  d ial 
    p ulse rec og nition as two method s for c allers to p rovid e non-sp oken input to the 
    system.
    Touch-Tone Recognition
    Touc h-tone rec og nition is a c ommon feature of voic e resp onse ap p lic ations.  The 
    majority of the telep hones in the United  States are eq uip p ed  with touc h-tone 
    servic e, b ut touc h-tone telep hone availab ility will b e d ifferent based  on loc ations.  
    A g reater p roportion of b usiness loc ations have touc h-tone c ap ab ilities than d o 
    resid enc es.  If your c allers have touc h-tone telep hones, it is ec onomic al and  
    effic ient to allow touc h-tone c aller inp ut for most transac tions.
    Touch-Tone Recognition Uses
    Touc h-tone inp ut may b e used  to selec t c hoic es from a sp oken menu.  It may 
    also b e used  to enter numeric al d ata suc h as c red it c ard  numb ers or p ersonal 
    id entific ation numb ers.
    Touch-Tone Recognition Capabilities
    Touc h-tone rec og nition rec og nizes the d ig its 0 throug h 9, as well as the asterisk 
    (*) and pound sign (#).
    With touc h-tone rec og nition, c allers have the op tion to resp ond  with a tone while 
    a prompt is playing.  This capability is known as 
    dial through (may also b e 
    referred  to 
    talkoff).  As soon as the system d etec ts the tone, the p romp t stops.
    The system also sup p orts 
    d ial ahead.  This c ap ab ility allows the system to c ollec t 
    touc h-tone inp ut as it is entered  b y c allers, even b efore they are p romp ted  for it.  
    The touc h-tone input is then used  in the ord er in whic h it was rec eived .  This 
    allows c allers to resp ond  to more than one p romp t at a time, without having  to 
    listen to the intermed iate p romp ts.  If exp erienc ed  c allers are familiar with the 
    up c oming  p romp ts, they d o not have to wait until the next p romp t starts p laying 
    b efore p ressing  the req uired  touc h-tone b uttons.
    Tone-touc h rec og nition c an b e used  in c onjunc tion with other rec og nition 
    method s, suc h as the Dial Pulse Rec og nition (DPR) software p ac kag e, as well as 
    the two sp oken-inp ut rec og nition software p ac kag es (WholeWord  and  
    FlexWord ™ sp eec h rec og nition) d isc ussed  later in this c hapter. 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Voice Response Advanced Technologies 
    Page 2-3 Touch-Tone and Dial Pulse Recognition 
    2
    Touch-Tone Recognition Accuracy
    The INTUITY CONVERSANT system is very ac c urate when rec og nizing  
    touc h-tone input.  However, c allers may make mistakes when entering 
    touc h-tones, and  a well d esigned  ap p lic ation must hand le c aller errors 
    g rac efully.
    Dial Pulse Recognition
    If touc h-tone servic e is not wid ely availab le in your area or c ountry of interest, you 
    c an offer d ial p ulse rec og nition to p rovid e non-sp oken inp ut to the system.
    Dial Pulse Recognition Uses
    The DPR software allows c allers with rotary telep hones, or p ush-b utton 
    telep hones that g enerate d ial pulses, to interac t with system ap p lic ations.  Muc h 
    like tone-touc h inp ut, d ial p ulse inp ut may used  to enter menu c hoic es and  
    numeric al d ata suc h as a b ank ac c ount numb er.
    Dial Pulse Recognition Capabilities
    The DPR software rec og nizes the d igits 0 throug h 9, but does not rec ognize the 
    asterisk (*) or p ound  sig n (#).
    DPR c an b e used  tog ether in an ap plic ation with touc h-tone rec og nition and  with 
    sp eec h rec og nition (WholeWord  or FlexWord ).
    Dial ahead and dial through are not supported for DPR.
    Dial Pulse Recognition Accuracy
    DPR is slig htly less ac c urate than touc h-tone rec og nition, but it d oes allow c allers 
    to interac t with the ap p lic ation without talking  to an ag ent.  The ac c urac y of DPR 
    c an b e imp roved , however, b y the use of sp ec ific  ap p lic ation d esig n tec hniq ues.  
    These tec hniq ues are disc ussed  in g reater d etail in Chap ter 4, ‘‘Desig ning  a 
    Voic e Resp onse Ap p lic ation.’’ 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Voice Response Advanced Technologies 
    Page 2-4 Speech Recognition 
    2
    Speech Recognition
    It makes g ood  b usiness sense to p rovid e the c ap ab ilities offered  b y the sp eec h 
    rec ognition software.  Sp eec h rec og nition allows your c allers to:
    nSp eak their resp onses; a more natural interfac e for c aller inp ut.
    nProvid e inp ut if they d o not have touc h-tone servic e.
    nProvid e inp ut for some typ es of information, like names, that d o not have a 
    natural translation to touc h-tone inp ut.
    WholeWord Speech Recognition
    The WholeWord  Sp eec h Rec og nition software p ac kag e is used to rec og nize 
    sp oken inp ut of c onnec ted  d ig its and  yes/no resp onses.
    WholeWord Speech Recognition Uses
    WholeWord  speec h rec og nition is most suc c essful when it is used  to aug ment a 
    touc h-tone ap p lic ation to p roc ess c allers who d o not have touc h-tone 
    telep hones.
    The b est app lic ations first ask c allers to ind ic ate whether they have a touc h-tone 
    telep hone (usually b y p ressing  one on the keyp ad ).  If no tone is d etec ted , the 
    ap plic ation p romp ts c allers to resp ond with sp oken inp ut (instead  of transferring 
    the c all to an attend ant).  In this manner, c allers who want to p rovid e sp oken 
    inp ut c an b e served  b y the system, instead  of req uiring  an attend ant.
    If your ap p lic ation req uires inp ut that c an easily b e map p ed  to touc h-tone 
    sig nals, d o not ig nore touc h-tone input in favor of sp eec h rec og nition.  For long er 
    d ig it seq uenc es, touc h-tone inp ut is more ac c urate.
    WholeWord  and FlexWord  sp eec h rec og nition c an b e used  in the same 
    ap plic ation for inc reased  flexib ility.  When d evelop ing  your ap p lic ation, you c an 
    sp ec ify a Promp t & Collec t ac tion to use either WholeWord or FlexWord  sp eec h 
    rec ognition, d ep end ing  on what you want the c aller to say.  For examp le, you c an 
    first p rompt “ Please say your ac c ount numb er.”   After the selec tion, the next 
    p romp t is “ Choose from the following  transac tions.  Say ‘ac c ount balanc e,’ or say 
    ‘transfer,’ or say ‘attend ant’ to sp eak to a servic e rep resentative.”
    Se e  ‘‘FlexWord  Sp eec h Rec og nition’’
     b elow for more information on FlexWord  
    sp eec h rec og nition. 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Voice Response Advanced Technologies 
    Page 2-5 Speech Recognition 
    2
    WholeWord Speech Recognition Capabilities
    This section describes the different WholeWord speech recognition capabilities.
    WholeWord Speech Recognition Languages
    The WholeWord  Sp eec h Rec og nition software is used  to rec og nize word s in the 
    following  lang uag es:
    nAustralian Eng lish
    nBrazilian Portug uese
    nCanad ian Frenc h
    nCastilian Sp anish
    nDutch
    nFren c h
    nGerman
    nJap anese
    nLatin-Americ an Sp anish
    nUK English
    nUS English
    Eac h lang uag e c an rec ognize:
    nEq uivalents for “ yes”  and  “ no”
    nSing le d ig its (zero throug h nine) and  c ommonly used  synonyms
    nA series of digits (also known as connected digits)
    Bilingual Applications
    Any two of the lang uag es listed  ab ove c an b e used  tog ether on a sing le system 
    to sup p ort biling ual ap plic ations.
    A b iling ual p erson c an rec og nize two lang uag es simultaneously.  A b iling ual 
    ap plic ation, however, c an only rec ognize sp eec h in a sing le lang uag e at any one 
    Promp t & Collec t ac tion.  You c an d esig n an ap plic ation that asks c allers to 
    ind ic ate their p referred  lang uag e, then play p romp ts and  announc ements, and 
    rec ognize the word s “ yes”  and  “ no”  as well as sp oken d ig its in one of the two 
    lang uag es installed  in your system.  This allows your system to und erstand  
    c allers who resp ond  in either lang uage you make availab le.  See ‘‘Biling ual, 
    Multilingual, and Non-US English Applications’’ in Chapter 4, ‘‘Designing a Voice 
    Resp onse Ap p lic ation,’’ for more information. 
    						
    							INTUITY CONVERSANT System Version 6.0
    Application Design Guidelines  
    585-310-670  Issue 1.0
    December 1996
    Voice Response Advanced Technologies 
    Page 2-6 Speech Recognition 
    2
    WholeWord Speech Recognition and Key Word Spotting
    WholeWord  speech rec og nition also sup p orts key word  sp otting .  A key word  is 
    one of the word s in a list of word s that the system is instruc ted  to rec og nize at a 
    p artic ular p oint in the transac tion.  Key word  sp otting  is the ab ility of the 
    rec og nizer to isolate a key word  from a stream of c aller input, inc lud ing  
    extraneous sp eec h, b ac kg round  noise, or c aller noises.  For this reason, c allers 
    d o not have to say the key word  in isolation.  For examp le, if the rec og nizer is 
    listening for c allers to say “ yes”  or “ no,”  it c an also rec og nize “ yes”  if c allers say 
    “ Yes, I d o.”   However, the rec og nizer find s a key word  most ac c urately when it is 
    said  alone, without any other word s or other noises b efore or after it.
    WholeWord Speech Recognition and Barge-in
    WholeWord speech recognition supports barge-in.  Experienced callers do not 
    have to wait until the end  of a promp t to b eg in sp eaking  their resp onses.  As 
    soon as the system rec og nizes something the c aller says, the p romp t stop s 
    playing.  This allows a single application to support both inexperienced and 
    exp erienc ed  c allers.
    WholeWord Speech Recognition Accuracy
    Dec ide c arefully where to allow sp oken inp ut.  Sp oken input is not the most 
    ap prop riate inp ut for all ap p lic ations.  Touc h-tone inp ut may b e faster and  more 
    ac c urate if your c allers are often sp eaking  from a noisy environment suc h as a 
    c ar or an airp ort.  Some c allers may have sec urity c onc erns ab out speaking 
    p rivate information (like an ac c ount numb er) aloud  if they use your servic e from 
    outsid e their homes.  If c allers must enter a long  series of d ig its or several d ata 
    items, touc h-tone inp ut may ac hieve b etter ac c urac y than WholeWord sp eec h 
    rec ognition.
    FlexWord Speech Recognition
    Another way for the system to proc ess sp oken inp ut is with the FlexWord  Sp eec h 
    Rec og nition software p ac kag e.  FlexWord  sp eec h rec og nition is used  to 
    rec ognize sp oken word s from a spec ific  set of word s, a 
    voc ab ulary, that you, the 
    ap plic ation d esig ner, d efine.  Allowing  c allers to say the op tion they want instead  
    of saying a number assigned to the option can make the interaction more natural 
    and  easier to use.
    FlexWord Speech Recognition Uses
    FlexWord  sp eec h rec og nition p rovides an intuitive, natural interfac e to the c allers 
    and  may be most suc c essful in ap p lic ations where it would  b e awkward  or 
    inc onvenient for c allers to enter touc h-tone input, suc h as when entering  a name.  
    FlexWord  sp eec h rec og nition ap p lic ations c an sup p ort more menu c hoic es than 
    either touc h-tone inp ut or WholeWord  sp eec h rec og nition.  Menus c an rang e 
    from small, suc h as a c hoic e of c lothing  sizes, up  throug h larg e, suc h as the 
    names of all 500 p eop le in your c omp any. 
    						
    All Lucent Technologies manuals Comments (0)

    Related Manuals for Lucent Technologies INTUITY CONVERSANT System Version 6.0 Guide