Contents

Keyword definitions

This topic provides descriptions of the various methods you can use to create keywords for use in Interaction Analyzer.

Keyword entry
Keyword user-defined pronunciations
keyword anti-spellings

Keyword entry

You enter keywords, up to 120 characters, through Interaction Administrator or the Interaction Administrator interface of IC Server Manager. When entering a keyword or phrase, you enter it with its normal spelling. The speech recognition engine uses that spelling to match it to its usual phonetic pronunciation.

Keyword spellings

You can add multiple spellings of a keyword, depending on the different ways in which a person pronounces a word, such as data. For this word, some people pronounce it as dayta while others pronounce it as dahta. The dictionary through which Interaction Analyzer compares pronunciations already handles many words with multiple pronunciations. However, if you enter terms that are specific to a specialized field, such as medicine or science, you may need to provide alternate spellings to ensure that Interaction Analyzer can recognize the word.

Keyword synonyms

One method to avoid regarding spellings is that of synonyms. For example, if you wanted a keyword of home, resist the temptation to enter the following synonyms as alternate spellings:

House
Abode
Domicile
Casa
Hacienda
Chateau
Castle
Shack
Hut
Home
Condominium
Apartment
Dwelling

Genesys recommends you do not enter synonyms as alternate spellings because Interaction Analyzer determines the spotability factor by the lowest, least discernable spelling. In the previous example, the lowest, least discernable spelling would be "Hut." Instead, create a keyword set titled home and enter each synonym as a keyword. This method ensures each synonym keyword has a separate spotability factor and confidence threshold, and can be further discerned through additional spellings and user-defined pronunciations.

Keyword acronyms

If you want to enter an acronym as a keyword, enter the acronym in capital letters. The following examples are acronyms that you can enter: USA, FBI, ID, ETA, EU, COD, HTTP, FTP, and so on. This method does not work for acronyms that are pronounced as words, such as SCSI (scuzzy), RAM, CD-ROM, PIN, NASA, and SCUBA.

Keyword punctuation

Interaction Analyzer ignores any punctuation characters that you enter in a keyword. These characters include commas, periods, exclamation points, question marks, colons, semicolons, special characters ($, %, &, ( ), @), and quotation marks.

Keyword numbers

Any numbers you enter as keywords are recognized as words. For example, if you enter H20, Interaction Analyzer recognizes this keyword as "H two O." However, if you enter multiple digits, Interaction Analyzer recognizes each number separately. For example, if you enter 100, Interaction Analyzer recognizes this keyword as "one, zero, zero," not "one-hundred." Likewise "21" is identified as "two, one," not "twenty-one."

Keyword contractions

Interaction Analyzer recognizes contractions when you enter them as keywords. Words such as don't, can't, won't, couldn't, it's, and I'll are acceptable.

Keyword abbreviations

Interaction Analyzer does not recognize abbreviations, such as Dr., Mrs., Mr., Jr., Sr., and others.

Keyword user-defined pronunciations

When you define keywords, Interaction Analyzer also allows you to enter phonetic, user-defined spellings based on ARPAbet. ARPAbet was a project that used ASCII characters to define the phonemes that make up all possible sounds in the English (US) language.

There are many words in English (US) that are spelled differently than the associated pronunciation. For example, tough, cough, dough, through, and bough are all similar in their spellings ("ough"), but have different pronunciations. Using user-defined pronunciations enables Interaction Analyzer to identify these words correctly. For example, the following table displays how these words are specified using ARPAbet phonemes:

Word	ARPAbet pronunciation
tough	t ah f
cough	c ao f
dough	d ow
through	th r uw
bough	b aw

The following table provides the ARPAbet phonemes and example words:

ARPAbet phoneme	Word examples	ARPAbet phoneme	Word examples
Vowels		Consonants
aa	father, cot	b	buy
ae	at, fast	ch	chair
ah	but, sun	d	day
ao	off, fall, frost	dh	that, the, them
aw	how, now	f	for
ay	my, why, ride	g	go
eh	red, men	hh	house
er	her, bird, hurt, nurse	jh	just
ey	say, eight	k	key
ih	big, win	l	late
iy	bee, she	m	man
ow	show, coat	n	non
oy	boy, toy	ng	sing
uh	should, could	p	pay
uw	you, new, food	r	run
y	yes	s	say
		sh	show
		t	take
		th	thanks, Thursday
		v	very
		w	way
		z	zoo
		zh	measure, pleasure

The following table provides examples of user-defined spellings of words in the English (US) language:

Word	User-defined spelling
Abdominal	ae b d aa m ah n ah l
Acrimonious	ae k r ah m ow n iy ah s
Affluent	ae f l uw ah n t
Articulated	aa r t ih k y ah l ey t ah d
Belaboring	b ih l ey b er ih ng
Botanists	b aa t ah n ih s t s
Channeling	ch ae n ah l ih ng
Charismatic	k eh r ih z m ae t ih k
Circumstantially	s er k ah m s t ae n ch ah l iy
Deformation	d iy f ao r m ey sh ah n
Domestication	d ah m eh s t ah k ey sh ah n
Eyeglasses	ay g l ae s ah z
Galvanizing	g ae l v ah n ay z ih ng
Hostilities	hh aa s t ih l ah t iy z
Infrastructure	ih n f r ah s t r ah k ch er

Another resource is the ARPAbet topic on Wikipedia at the following URL address: http://en.wikipedia.org/wiki/Arpabet.

User-defined pronunciations are also useful for the different dialects you can encounter in a language. For example, consider the English (US) word "lawyer" (l ao y er). Someone from Massachusetts may pronounce the word as "lahyah" (l aa y ah) while someone from Alabama could pronounce it "lawryer" (l ao r y uh r). User-defined spellings can help you define the keyword so Interaction Analyzer can correctly produce a match.

Interaction Analyzer uses extensions to ARPAbet for languages that need more phonemes to cover the phonology of the language.

Important: There are words in the medical or scientific fields that are based on Latin and Greek, but have had the associated pronunciations transformed over centuries of use, such as peritonitis. For these words, if Interaction Analyzer has difficulty identifying the keyword based on the correct spelling, Genesys recommends you enter user-defined pronunciations to ensure proper identification of those unique keywords.

Keyword anti-spellings

After you define keywords, you have the option of entering anti-spellings that specify similar-sounding words you do not want mistaken for that keyword. These are words that could confuse Interaction Analyzer and be marked as instances of that keyword, such as false positives.

For example, the following table provides some keywords and possible anti-spellings you can add to the definitions:

Keyword	Anti-keywords
Lawyer	Lower Loiter Foyer Employer
Surely	Charlie Surly Purely Journey
Guarantee	Warranty Guillotine Green tea
Unfair	Conveyor Unveil Affair A fare On there

Additionally, you can specify anti-spellings where the keyword is part of a larger word. For example, consider the word form. You probably would not want Interaction Analyzer to spot this word when it is a part of larger words, such as uniform, formatted, and formation. In this case, if Interaction Analyzer does spot the keyword within the larger words, you can enter the mistaken words as anti-spellings.

Do not try to determine anti-spellings when you define a keyword. Instead, when your keywords are active and are run against real interactions in the contact center, you can analyze recordings, identify words mistaken as keywords, and add them as anti-spellings to the keyword definition.

If you eventually enter anti-spellings in a keyword definition, you can also consult numerous Web sites that list other rhyming words for that specific keyword.