Checking pronunciations with dicttest

The dicttest program is a command line tool that lets you check pronunciations of individual words or lists of words inside a file. The program is installed with Recognizer.

Note: The following information is written with English examples, but applies generically to all languages. The English examples use English phonemes, but you must always use the phonemes for the language you specify (again, see the appropriate Language Supplement for a list of phonemes).

Program usage

Use a command prompt window to gain access to the program. For example, typing the program name with no arguments provides usage information: 

c:/> dicttest
Usage: dicttest -language code
  [-system_dict file]
  [-user_dict file] 
  [-max_pron num]
  [-encoding]

Argument

Meaning

-language code

Required. Specifies the ID of the language being tested. For example, to test French you would enter "-language fr-FR" or "-language fr-CA".

-system_dict file

Optional (required if -language is omitted). Specifies a system dictionary to search for pronunciations. System dictionaries are binary files supplied with Recognizer; there is one system dictionary per language stored in the installation baseline. This parameter is useful when using a system dictionary that is not stored in the default baseline location.

-user_dict file

Optional. Specifies a user dictionary to search for pronunciations. If a word is found in the user dictionary, the program uses those pronunciations and does not search the system dictionary. The precedence of the file can be changed (see Changing the default precedence) by appending the SWI.type variable; for example "user.dict?SWI.type=backup".

-max_pron num

Optional. Specifies the maximum number of pronunciations to be shown for any given word. By default, dicttest shows all pronunciations in the available dictionaries.

-encoding

Optional. Specifies the encoding of the dictionary file. Values are -utf8 (default) or -iso8859.

See dicttest.

When the program starts, you can type one word at a time. The program finds all relevant pronunciations for the word you type. Example (for US English): 

c:/> dicttest -language en-US -system_dict "dict_path"
enter word: sally
Pron 0: s{li:
Source: sally?name=system;level=backup;massaged=0;
enter word: ^C

To stop the program, press CTRL/C.

Above, the value of dict_path is:

%SWISRSDK%\config\en-US\dictionary\Basic.dictionary

Program output

The dicttest program provides the following output:

  • Count of the number of pronunciations found.
  • The phonemic spelling of the word (the pronunciation).
  • Indication of the location where the pronunciations were found.
  • A max_res_len value (used internally by Nuance).

Here is sample output for the word "hello":

Pron 0: h@l@U
Source: hello?name=system;level=backup;massaged=0;
 
Pron 1: hel@U
Source: hello?name=system;level=backup;massaged=0;
 
Pron 2: hVl@U
Source: hello?name=system;level=backup;massaged=0;

Explanation of the output:

Output

Description

Pron 0/1/2

Three pronunciations were found.

name=system

All pronunciations were found in the system dictionary.

level=backup

Indicates the precedence of the named dictionary. See Dictionary precedence. (The system dictionary is a backup dictionary).

massaged=0

Boolean that indicates whether the lookup algorithm employed "mangling" rules. For example, if the algorithm tried variations of the exact word after the removal of punctuation.

Automatically generated pronunciations

As mentioned earlier (The system dictionary), if you enter a word that is not found in a dictionary, the system automatically generates a pronunciation and indicates that a set of internal rules were used as the source of the pronunciation. Some languages may write an additional warning message to Recognizer’s diagnostic log (by default, this log file is $NUANCE_DATA_DIR/system/diagnosticLogs/nrs.log.

Investigate generated pronunciations carefully to ensure that they are correct. For example, if a person’s name "Pnina" returned a pronunciation that is not correct, and the application developer would make a correction (in this case, the pronunciation is correct): 

pni:n@
source: pnina?name=automatic;level=automatic;massaged=1;

Explanation of the output:

Output

Description

name=automatic

The "automatic" dictionary was used to generate the pronunciation.

level=automatic

The precedence of the automatic dictionary is also "automatic". See Dictionary precedence. (When the automatic dictionary is used, the precedence is always "automatic" too.)

massaged=1

Boolean that indicates whether the lookup algorithm employed "mangling" rules. For example, if the algorithm tried variations of the exact word after the removal of punctuation.

Use the indirection "<" operator to send an input file to dicttest. For example, type a file of simple words: 

c:/> type t.txt
dog
cat
jean
pnina

Use the file as input:

c:/> dicttest -language en-US 
  -system_dict"%SWISRSDK%\Basic.dictionary" < t.txt
 
enter word:
 
...output not shown
 

The output would show the pronunciations of each word in the list.

Below is a sample user dictionary that defines a pronunciation for a phrase that is formed by joining words with the underscore character (_): 

<?xml version="1.0" encoding="UTF-8" ?>
<lexicon xml:lang="en-us">
 <entry key="at_the_market">
  <definition value="{t#sil#D@#sil#mQrk@t" />
  <definition value="{t#sil#Di:#sil#mQrk@t" />
 </entry>
</lexicon>

Using this dictionary, here is a sample session with the dicttest tool:

dicttest -language en-us -user_dict atthemarket.xml
Word: at the market
 
Pron 0: {t#sil#D@#sil#mQrk@t
Source: at?name=system;level=backup;massaged=0;the?
name=system;level=backup;massaged=0;market?name=system;
level=backup;massaged=0;
 
Pron 1: {t#sil#Di:#sil#mQrk@t
Source: at?name=system;level=backup;massaged=0;the?
name=system;level=backup;massaged=0;market?name=system;
level=backup;massaged=0;
 
Pron 2: @t#sil#D@#sil#mQrk@t
Source: at?name=system;level=backup;massaged=0;the?
name=system;level=backup;massaged=0;market?name=system;
level=backup;massaged=0;
 
...more pronunciation variations...
 
enter word: at_the_market
 
Pron 0: {t#sil#D@#sil#mQrk@t
Source: at_the_market?name=user;level=primary;massaged=0;
 
Pron 1: {t#sil#Di:#sil#mQrk@t
Source: at_the_market?name=user;level=primary;massaged=0;