Phoneme Tables

• Phoneme files
• Phoneme Definitions
• Phoneme Properties
  • Type
  • Properties
  • Place of Articulation
  • starttype
  • endtype
  • lengthmod
  • voicingswitch
• Phoneme Instructions
  • length
  • ipa
  • WAV
  • FMT
  • VowelStart
  • VowelEnding
  • Vowelin
  • Vowelout
  • ChangePhoneme
  • ChangeIfDiminished
  • ChangeIfUnstressed
  • ChangeIfNotStressed
  • ChangeIfStressed
  • IfNextVowelAppend
  • RETURN
  • CALL
• Conditional Statements
  • Conditions
  • Attributes
• Sound Specifications
• Vowel Transitions

A phoneme table defines all the phonemes which are used by a language, together with their properties and the data for their production as sounds.

Generally each language has its own phoneme table, although additional phoneme tables can be used for different voices within the language. These alternatives are referenced from Voice files.

A phoneme table does not need to define all the phonemes used by a language. It can inherit the phonemes from a previously defined phoneme table. For example, a phoneme table may redefine (or add) some of the vowels that it uses, but inherit most of its consonants from a standard set.

The source files for the phoneme data are in the "phsource" directory in the espeakedit download package. "Vowel files", which are referenced in FMT(), VowelStart(), and VowelEnding() instructions are made using the espeakedit program.

Phoneme Files

The phoneme tables are defined in a master phoneme file, named phonemes. This starts with the base phoneme table followed by phoneme tables for other languages and voices. These inherit phonemes from the base table or previously defined tables.

In addition to phoneme definitions, the phoneme file can contain the following:

include <filename>

Includes the text of the specified file at this point. This allows different phoneme tables to be kept in different text files, for convenience. is a relative path. The included file can itself contain include statements.

phonemetable <name> <parent>

Starts a new phoneme table, and ends the previous table.

Is the name of this phoneme table. This name is used in Voice files.

Is the name of a previously defined phoneme table whose phoneme definitions are inherited by this one. The name base indicates the first (base) phoneme table.

Phoneme Definitions

A phoneme table contains a list of phoneme definitions. Each starts with the keyword phoneme and the phoneme name (this is the name used in the pronunciation rules in a language's *_rules and *_list files), and ends with the keyword endphoneme.

The phoneme mnemonics are based on the scheme by Evan Kirshenbaum which represents International Phonetic Alphabet symbols using ascii characters.

For example:

phoneme aI
  vowel
  starttype #a endtype #i
  length 230
  FMT(vowels/ai)
endphoneme

phoneme s
  vls alv frc sibilant
  voicingswitch z
  lengthmod 3
  Vowelin  f1=0  f2=1700 -300 300  f3=-100 80
  Vowelout f1=0  f2=1700 -300 250  f3=-100 80  rms=20

  IF nextPh(isPause) THEN
    WAV(ufric/s_)
  ELIF nextPh(p) OR nextPh(t) OR nextPh(k) THEN
    WAV(ufric/s!)
  ENDIF
  WAV(ufric/s)
endphoneme

A phoneme definition contains both static properties and executed instructions. The instructions may contain conditional statements, so that the effect of the phoneme may be different depending on adjacent phonemes, whether the syllable is stressed, etc.

The instructions of a phoneme are interpreted in two different phases. In the first phase, the instructions may change the phoneme and replace it by a different phoneme. In the second phase, instructions are used to produce the sound for the phoneme.

The import_phoneme statement can be used to copy a previously defined phoneme from a specified phoneme table. For example:

phoneme t
  import_phoneme base/t[
endphoneme

means: phoneme t in this phoneme table is a copy ofphoneme t[ from phoneme table base. A length instruction can be used after import\_phoneme to vary the length from the original.

Phoneme Properties

Within the phoneme definition the following lines may occur: ((V) indicates only for vowels, (C) only for consonants).

Type

One of these must be present.

Properties

Place of Articulation

starttype

starttype <phoneme>

Allocates this phoneme to a group so that conditions such as nextPh(#e) can test for any of a group of phonemes. Pre-defined groups for use for vowels are: #@ #a #e #i #o #u. Additional groups can be defined as phonemes with type virtual.

endtype

endtype <phoneme>

Allocates this phoneme to a group so that conditions such as prevPh(#e) can test for any of a group of phonemes. Pre-defined groups for use for vowels are: #@ #a #e #i #o #u. Additional groups can be defined as phonemes with type virtual.

lengthmod

lengthmod <integer>

(C) Determines how this consonant affects the length of the previous vowel.

This value is used as index into the length_mods table in the CalcLengths() function in the eSpeak NG program.

voicingswitch

voicingswitch <phoneme>

This is used for some languages to change between voiced and unvoiced phonemes.

Phoneme Instructions

Phoneme Instructions may be included within conditional statements.

During the first phase of phoneme interpretation, an instruction which causes a change to a different phoneme will terminate the instructions. During the second phase, FMT() and WAV() instructions will terminate the instructions.

length

length <length>

The relative length of the phoneme in miliseconds. Typical values are about 140 for a short vowel and from 200 to 300 for a long vowel or diphong. A length() instruction is needed for vowels. It is optional for consonants.

ipa

ipa <ipa string>

In many cases, eSpeak NG makes IPA (International Phonetic Alpbabet) phoneme names automatically from eSpeak NG phoneme names. If this is not correct, then the phoneme definition can include an ipa instruction to specify the correct IPA name. IPA strings may include non-ascii characters. They may also include characters specified by their character codes in the form U+ followed by 4 hexadecimal digits. For example a string: aU+0303 indicates 'a' with a 'combining tilde'.

WAV

WAV(<wav file>, <amplitude>)

is a path to a WAV file (22 kHz, 16 bits, mono) within phsource/ which will be played to produce the sound. This method is used for unvoiced consonants. does not include a .WAV filename extension, although the file to which it refers may or may not have one.

is optional. It is a percentage change to the amplitude of the WAV file. So, {gfm-extraction-a2c65f643e78c2fc3b0fd06e70d10395} means: play file 'ufric/s.wav' at 50% amplitude. Default value is 100.

FMT

FMT(<vowel file>, <amplitude>)

is a path to a file (within phsource/) which defines how to generate the sound (a vowel or voiced consonant) from a sequence of formant values. Vowel files are made using the espeakedit program which is not part of this project.

is optional. It is a percentage change to the amplitude of the sound which is synthesized from the {gfm-extraction-b5a7ac9ff7d0d019e397a298d853dcbc} instruction. Default value is 100.

For voiced consonants, a {gfm-extraction-b5a7ac9ff7d0d019e397a298d853dcbc} instruction may be followed by an {gfm-extraction-ec46da7ce21a597f548ea3f14576f643} instruction. {gfm-extraction-ec46da7ce21a597f548ea3f14576f643} has the same format as a {gfm-extraction-868963665678b6340907f6e82b096272} instruction, but the WAV file is mixed with the sound which is synthesized from the {gfm-extraction-b5a7ac9ff7d0d019e397a298d853dcbc} instruction.

VowelStart

VowelStart(<vowel file>, <length adjust>)

This is used to modify the start of a vowel when it follows after a sonorant consonant (such as [l] or [j]). It replaces the first frame of the which is specified in a FMT() instruction by this , and adjusts the length of the original by a signed value . The VowelStart() instruction may be specified either in the phoneme definition of the vowel, or in the phoneme definition of the sonorant consonant which precedes the vowel. The former takes precedence.

VowelEnding

VowelEnding(<vowel file>, <length adjust>)

This is used to modify the end of a vowel when it is followed by a sonorant consonant (such as [l] or [j]). It is appended to the which is specified in a FMT() instruction by this , and adjusts the length of the original by a signed value . The VowelEnding() instruction may be specified either in the phoneme definition of the vowel, or in the phoneme definition of the sonorant consonant which follows the vowel. The former takes precedence.

Vowelin

Vowelin <vowel transition data>

(C) Specifies the effects of this consonant on the formants of a following vowel. See vowel transitions.

Vowelout

Vowelout <vowel transition data>

(C) Specifies the effects of this consonant on the formants of a preceding vowel. See vowel transitions.

ChangePhoneme

ChangePhoneme(<phoneme>)

Change to the specified phoneme.

ChangeIfDiminished

ChangeIfDiminished(<phoneme>)

Change to the specified phoneme (such as schwa, @) if this syllable has "diminished" stress.

ChangeIfUnstressed

ChangeIfUnstressed(<phoneme>)

Change to the specified phoneme if this syllable has "diminished" or "unstressed" stress.

ChangeIfNotStressed

ChangeIfNotStressed(<phoneme>)

Change to the specified phoneme if this syllable does not have "primary" stress.

ChangeIfStressed

ChangeIfStressed(<phoneme>)

Change to the specified phoneme if this syllable has "primary" stress.

IfNextVowelAppend

IfNextVowelAppend(<phoneme>)

If the following phoneme is a vowel then this additional phoneme will be inserted before it. Usually it is short pause to distinguish two vowels from diphthongs.

RETURN

Ends executions of instructions.

CALL

CALL <phoneme table>/<phoneme>

Executes the instructions of the specified phoneme.

Conditional Statements

Phoneme definitions can contain conditional statements such as:

IF <condition> THEN
    <statements>
ENDIF

or more generally:

IF <condition> THEN
    <statements>
ELIF <condition> THEN
    <statements>
...
ELSE
    <statements>
ENDIF

where the ELSE and multiple ELSE parts are optional.

Multiple conditions may be joined with AND or OR, but not a mixture of ANDs and ORs.

A condition may be preceded by NOT. For example:

IF <condition> AND NOT <condition> THEN
    <statements>
ENDIF

Conditions

Conditions can be:

thisPh(<attribute>)

Test this current phoneme

prevPh(<attribute>)

Test the previous phoneme

prevPhW(<attribute>)

Test the previous phoneme, but only within the same word. Returns false if there is no previous phoneme in the word.

prev2PhW(<attribute>)

Test the phoneme before the previous phoneme, but only within the same word. Returns false if it is not in this word.

nextPh(<attribute>)

Test the following phoneme

next2Ph(<attribute>)

Test the phoneme after the next phoneme.

nextPhW(<attribute>)

Test the next phoneme, but only within the same word. Returns false if there is no following phoneme in the word.

next2PhW(<attribute>)

Test the phoneme after the next phoneme, but only within the same word. Returns false if not found before the word end.

next3PhW(<attribute>)

Test the third phoneme after the current phoneme, but only within the same word. Returns false if not found before the word end.

nextVowel(<attribute>)

Test the next vowel after the current phoneme, but only within the same word. Returns false if there is none.

prevVowel(<attribute>)

Test the previous vowel before the current phoneme, but only within the same word. Returns false if there is none.

PreVoicing()

This is used as part of the instructions for voiced stop consonants (e.g. [d] and [g]). If true then produce a voiced murmur before the stop.

KlattSynth()

Returns true if the voice is using the Klatt synthesizer rather than the eSpeak synthesizer.

Attributes

<phoneme name>

True if the phoneme has this phoneme name.

<phoneme group>

True if the phoneme has this starttype (or if it has this endtype if it is used in prevPh()). The pre-defined phoneme groups are #@, #a, #e, #i, #o, #u.

isPause

True if the phoneme is a pause.

isPause2

nextPh(isPause2) is used to test whether the next phoneme is not a vowel or liquid consonant within the same word.

isVowel
isNotVowel
isLiquid
isNasal
isVFricative

These test the phoneme type.

isPalatal
isRhotic

These test whether the phoneme has this property.

isWordStart
notWordStart

These text whether this is the first phoneme in a word.

isWordEnd

True if this is the final phoneme in a word.

isFirstVowel
isSecondVowel
isFinalVowel

True if this is the First, Second, or Last vowel in a word.

isAfterStress

True if this phoneme is after the stressed vowel in a word.

isVoiced

True if this phoneme is a vowel or a voiced consonant.

isDiminished

True if the syllable stress is "diminished"

isUnstressed

True if the syllable stress is "diminished" or "unstressed"

isNotStressed

True if the syllable stress is not "primary stress".

isStressed

True if the syllable stress is "primary stress".

isMaxStress

True if this is the highest stressed syllable in the word.

Sound Specifications

There are three ways to produce sounds:

• Playing a WAV file, by using a WAV() instruction. This is used for unvoiced consonants such as [p], [t] and [s].
• Generating a wave from a sequence of formant parameters, by using a FMT() instruction. This is used for vowels and also for sonorants such as [l], [j] and [n].
• A mixture of these. A stored WAV file is mixed with a wave generated from formant parameters. Use a FMT() instruction followed by addWav(). This is used for voiced stops and fricatives such as [b], [g], [v] and [z].

Vowel Transitions

These specify how a consonant affects an adjacent vowel. A consonant may cause a transition in the vowel's formants as the mouth changes shape between the consonant and the vowel. The following attributes may be specified. Note that the maximum rate of change of formant frequencies is limited by the program.

len=<integer>

Nominal length of the transition in miliseconds. If omitted a default value is used.

rms=<integer>

Adjusts the amplitude of the vowel at the end of the transition. If omitted a default value is used.

f1=<integer>

0: f1 formant frequency unchanged.

1: f1 formant frequency decreases.

2: f1 formant frequency decreases more.

f2=<freq> <min> <max>

: The frequency towards which the f2 formant moves (Hz).

: Signed integer (Hz). The minimum f2 frequency change.

: Signed integer (Hz). The maximum f2 frequency change.

f3=<change> <amplitude>

: Signed integer (Hz). Frequency change of f3, f4, and f5 formants.

: Amplitude of the f3, f4, and f5 formants at the end of the transition. 100 = no change.

brk

Break. Do not merge the synthesized wave of the consonant into the vowel. This will produce a discontinuity in the formants.

rate

Allow a greater maximum rate of change of formant frequencies.

glstop

Indicates a glottal stop.