cws
Greetings Guest
home > library > journal > view_article
« Back to Articles » Journal
Pronunciation Estimation
2▲ 2 ▼ 0
This public article was written by [Deactivated User], and last updated on 19 Sep 2015, 06:29.

[comments] Menu 1. Using PhoMo 2. Example 3. Limitations As you may or may not know, dictionaries on CWS have two fields were you can enter X-SAMPA or IPA into a word's entry in order to give its pronunciation. However, should you not use these, or enter a pronunciation, CWS will attempt to guess the pronunciation of the word based on your language's phonology and orthography (if you have set these up).

The most important thing to making sure this works even remotely correctly, is to ensure that every letter used in your words has a sound assigned to it on the inventory page. If you have letters that have no sound assigned, you can use the Zero Consonant or Zero Vowel to add these to the list anyway. Any letters that do not have a sound assigned will be estimated as /*/.

So what happens when - for example - /b/ changes to /v/ when before front vowels? Or voiced consonants become devoiced when in a cluster with other voiceless consonants? This is where we can use Rewrite rules to change the estimation of the word. If you're new to PhoMo, there are three help articles that split up how to use it:

[PhoMo Page Help]
[Intro to PhoMo rules]
[PhoMo Examples Reference]

To enable the rewrite rules, and to create the special PhoMo ruleset for this, you need to visit the pronunciation estimation page (presently found also under the green menu). Once this has been created, and enabled, you can begin writing your rules on the PhoMo page.

[top]Using PhoMo

Using PhoMo can be a little daunting for the first time, but please have a skim read of the above articles if you need to familiarise yourself with it.

An important note when editing the rewrite rules is that it uses the scope 'IPA' for the categories. That means that any categories that you have created so far with the scope 'general' cannot be used. This is because those with 'general' are used for modify the orthography of a word, not the IPA. You'll need to create new categories with IPA characters to be able to use categories in your PhoMo rules. There are two buttons on the creation page that will auto-create a 'C' and 'V' category for you will consonants and vowels respectively.

In order to save the ruleset you have created, it needs to be put into the Ruleset field, and you must have at least one word in the 'Input' field. Use this as a means to test that the rules you have created work properly. When you then run PhoMo, it will save your ruleset.

[top]Example

Using the example two changes used above, we will show you an example of some rules that can be created to show the desired effect. Let's say we have the words: bimo and azten. The two changes we want to make listed earlier are:

  • /b/ becomes /v/ when before front vowels; and
  • voiced consonants are devoiced when adjacent other voiceless consonants


Based on the phonology and orthography of the language, bimo and azten would be estimated as /bimo/ and /azten/ respectively. However, these are not correct. Using the following rules, we can change this:

b/v/_F D/T/_T|T_

The above rules use a few categories. We are assuming here that 'F' are our front vowels, 'D' are voiced consonants, and 'T' are voiceless consonants. These rules will force the estimation of bimo and azten to become /vimo/ and /asten/ as they should be.

[top]Limitations

As with any computer program, there are limitations on how complex it can think. The CWS estimation for pronunciation:
  • can not predict stress
  • will always pick a digraph over a monograph
  • can not analyse beyond word boundaries


Stress
The reason PhoMo cannot predict stress is because stress is subjective to each language. This makes it generally unpredictable as PhoMo doesn't know what your language's phonotactics are like. For example, the word 'kaksen' could be split up a number of ways: 'ka-ksen', 'kak-sen', 'kaks-en', 'ka-kse-n', etc.

Some members on CWS have had some success with using PhoMo to insert the stress marker in IPA /./ into words using their languages regular rules (this is for example, particularly easy if your language has a strict CV syllable pattern). If you're able to do this, then you can use PhoMo rules to make changes that are relative to syllable boundaries.

Digraphs before monographs
Suppose a language has <g> as /g/, <h> as /h/, and <gh> as /ɣ/. Then we put the word nagheb through the estimation algorithm, and it will always spit out /naɣeb/. The only way around this is to use rewrite rules to determine when it should be /gh/ and when it should be /ɣ/.

Word boundaries
You cannot go beyond the word boundaries with the estimations to make that change. In other words, let's pretend you have the word(s) aner sepa and your language has it set up that <rs> is /ʂ/. The estimation will always come out as /anɛr sɛpa/ and never /anɛʂepa/. Since estimations are evaluated word-by-word, this sort of change cannot ever be reflected in the pronunciation estimation of the language.
Comments
[link] [quote] [move] [edit] [del] 14-Aug-20 03:21 [Deactivated User]
Please note that nobody is informed when people leave questions on system articles. If you need help, use the forums.
[link] [quote] 10-Aug-18 06:06 [Deactivated User]
No, but not at all attempting to modify the pronunciation would, because all your attempts to change it are just shoving random capital letters in for no reason. Capital letters are never IPA, so there is no circumstance in which they should be in the pronunciation.
[link] [quote] [move] [edit] [del] 10-Aug-18 05:39 [Deactivated User]
I used a online tool to genrate the glyphs and realised later that the letters were quite different to the sounds I wanted in English. Would assigning more similar letters resolve the problem?
[link] [quote] 10-Aug-18 05:34 [Deactivated User]
As far as I can tell, it worked just fine originally, then you attempted to fix it and broke it instead. It's hard to tell though because your orthography doesn't make a lot of sense, so it's possible you wanted it to be very... very... very different.
[link] [quote] [move] [edit] [del] 10-Aug-18 05:20 [Deactivated User]
can't seem to get this to work. i have assigned letters to sounds, but it does not seem to have worked.
[link] [quote] 08-Oct-17 19:03 [Deactivated User]
@LordoftheFrogs please post your query in the boards :)
[link] [quote] [move] [edit] [del] 08-Oct-17 17:01 [Deactivated User]
How do I make it turn an unaspirated stop into an aspirated one before a vowel when they are allophones?
privacy | FAQs | rules | statistics | graphs | donate | api (indev)
Viewing CWS in: English | Time now is 29-Mar-24 04:53 | Δt: 1315.407ms