break strings into syllables

Hello every one,

I have to write transcription for over 120 000 words in a worksheet “Text” in Column A. :-(
Now the ugliest part is to divide the transcription into syllables. And I need automation for that. Of course there will be a lot to correct afterwards but I hope that atomization and latter the correction will consume less time compared to doing all the work manually. I hope!
I have automated the conversion of special language letters to transcription letters in excel. Excel converts the strings to transcription itself and I have to make the corrections latter. I have experienced that I need a lot less time for this than typing whole the transcription myself.
I wonder if atomization of dividing the transcription into syllables and then manually correcting it will also save me that much time.
I guess that for atomization of breaking strings into syllables would need few pre defined criteria. First of all vowels and consonants have to be defined. For that purpose I have an extra worksheet “v&c”. Column A of this “v&c” contains of vowels and column B contains of consonants. For example ( the real consunants and vowels may be different. Its just an example here):
Consunants
b
c
d
f
g
h
j
k
l
m
n
p
q
r
s
t
v
w
x
z
ß
µ

Vowels
a
e
i
o
u
ä
ö
y


Secondly, the positions of consonants and vowels have to be analyzed. There are few cases possible:
Before:
love
ballade
transcription
country
abba
aba
abab
möhren
mässigung
schutz suchen
suchende person
combination of words

Should be:
lo-ve ( consunant vowel - consunant vowel )
bal-la-de ( con vow con - con vow -con vow )
tran-scrip-tion ( con con vow con -con con con vow con - con vow vow con )
coun-try ( con vow vow con - con con vow )
ab-ba ( vow cons - con vow )
a-ba ( vow - con vow )
a-bab ( vow - con vow con )
möh-ren ( con vow con - con vow con )
mäs-si-gung ( con vow con - con vow - con vow con con )
schutz suc-hen ( …. )
suc-hen-de per-son ( …. )
com-bi-na-tion of words ( …. )

I will be very thankful if someone could help me out with it.

To do this, you'd need a set of criteria for a syllable. Can you create this set? I'd think it an impossible task, especially if you are dealing with more than one language.

However, if you can come up with a set of criteria, I can definitely automate it for you.

Hi wali

Wow, this is more complicated than I had originally anticipated. Give this a shot and let me know if you can come up with any upgrades.

Please Login or Register  to view this content.

Hi,

i ran the macro over the example i have attached to the thread. It gives following results:

love .=> lo-ve
ballade .=> ba-l-laa-de (should be => bal-la-de)
transcription .=> tra-nscri-p-tii-on (should be => tran-scrip-tion)
country .=> co-untry (should be => coun-try)
abba .=> ab-baa (should be => ab-ba)
aba .=> aba (should be => a-ba)
abab .=> aba-b (should be => a-bab)
möhren .=> möhre-n (should be => möh-ren)
mässigung .=> mässi-gu-ng (should be => mäs-si-gung)
schutz suchen .=> schu-tz su-c-hee-n (should be => schutz suc-hen)
suchende person .=> su-c-hee-n-dee- -pee-r-soo-n (should be => suc-hen-de per-son)
combination of words .=> co-m-bii-na-ti-on- oo-f wo-rds (should be => com-bi-na-tion of words)
balabali .=> ba-la-ba-li- (should be => ba-la-ba-li)
abbi .=> ab-bii (should be => ab-bi)


The first thing is that the code doubles some vowels. Those vowels get double if they follow a double letter.

Secondly,how can we change it so that if a vowel is followed by two cons. and a vowel that the "-" is put between the cons?

It seems that it may really work. i hope that you can find time to make some adjustments.

I thank you a million time for taking the time to solve my problem. Thanks

I have fixed the vowel doubling.
In terms of your second request, can you give me a specific example?
Lastly, I don't know how to type the foreign letters (like german umlouts), but you can add them to the code yourself. Go to the Vowel Function, and add those letters to the Cases. I've added an example commented out. That should fix words like your example mohren.

Please Login or Register  to view this content.

Hi,

the double vowel is fixed. Great!!
Ok! i will change the vowels according to my needs in the code.

What do you think? can you also fix following things?:


original => your code returns now => but should be
ballade .=> ba-l-lade (.should be => bal-la-de ) -->no isolation of a cons like -l-
transcription .=> tra-nscri-p-tion (.should be => tran-scrip-tion ) --> at least one cons after vow before putting "-" , if vow is not the first letter of the word
country .=> co-untry (.should be => coun-try ) --> if two vowels follow each other both should be considered as one plus cons after vowel
aba .=> aba (.should be => a-ba )--> if word starts with vowel and third letter is vowel again then isolate the first vowel

abab .=> aba-b (.should be => a-bab )


thanks

You've got to remember we're dealing with set criteria. Just putting examples without a criteria does no good.
Original: I think you missed some text here.
ballade: Your first two criteria cause this. I can't fix it without messing up those criteria.
etc.

I told you when we started that this seemed to be an impossible task. Theoretically, you could add thousands of criteria to try and decypher where the "-" should be, but in the end there are simply too many possibilities. I think you'd be best off doing it manually, but you can use my code to try to build your own if you so desire.

What about ?

angst-schreeuw
bou-wen
ge-e-ve-naard
idee-ën
sba-gli-a-to
wij-zi-ging

Syllables in English are quite different from syllables in German, Dutch or Italian.
Even in 1 language there is no single algorithm.

From http://en.wikipedia.org/wiki/Syllabification

In some languages, the spoken syllables are also the basis of syllabification in writing. However, possibly due to the weak correspondence between sounds and letters in the spelling of modern English, written syllabification in English is based mostly on etymological or morphological instead of phonetic principles. For example, it is not possible to syllabify "learning" as lear-ning according to the correct syllabification of the living language. Seeing only lear- at the end of a line might mislead the reader into pronouncing the word incorrectly, as the digraph ea can hold many different values. The history of English orthography accounts for such phenomena.

English written syllabification therefore deals with a concept of "syllable" that doesn't correspond to the linguistic concept or a phonetic (as opposed to morphological) unit.

As a result, even most native English speakers are unable to syllabify words accurately without consulting a dictionary or using a word processor[citation needed]. The process is, in fact, so complicated that even schools usually do not provide much more advice on the topic than to consult a dictionary. In addition, there are differences between British and US syllabification and even between dictionaries of the same English variety.

In Finnish, Italian, and other nearly phonetically spelled languages, writers can in principle correctly syllabify any existing or newly created word using only general rules. In Finland, children are first taught to hyphenate every word until they produce the correct syllabification reliably, after which the hyphens can be omitted.

I have no idea where Pashto falls in all that.

It's imposible to divide original Pashto script (right to left) into syllables. Because, you dont have always visible vowels in the script. That means one word can be read in different ways and you have to read out from context which word it is. There are many examples where semantically you have four or five different words but they are written all with same spelling.

But I have here a transcription and not the original script. Secondly the words i have, their lenght dont exceed 12 Letters. No vowel is missing because the trascription is written in IPA(international phonetic alphabets). And I am quite sure that up to an extent the division into syllables is possible. Of course not 100% correctly but if enough critereas are defined and studied I am sure that up to 40% it may be divided correctly. I will try to studie the different possibe cases.

Of course for english it is not possible at all.

You could categorize words by vowel/consonant formation, e.g.,

youthfulness -> aaabbbabbabb
schmaltziest -> bbbbabbbaabb

There are 891 different such patterns among the 11,000-odd 12-letter words from the SOWPADS.

Wow!

What a fascinating thread. I agree with everyone, although you may be able to get some defined logic for English syllables, the number of possibilities are massive.
I've never heard of IPA before, but when comparing Japanese to English, Japanese would have to easier as I think it is one of the more (/most?) phonetic languages.

Dave,
Great work on putting the logic down in a working code
I'm not sure how fast the code runs now, but to make it faster, you could load the info into an "in memory array", work on array elements "in memory" & then write the results back to the spreadsheet in one hit (or fewer than currently). This minimises the "hits" on the spreadsheet. See here for some links or here for an alternative example.

Wali,
If you want to improve the speed of the code, you can "Order" the Select Case statements to have the characters listed from most likely to least likely and finally the "Case Else" as the last statement. This could be useful if you end up with a large list of other characters, because I believe a Select Case only processes the options until it finds the first True case statement. Here are a couple of links which discuss letter frequency (I'm guessing you may have something similar already?) that you could use to order your Case Statements:
http://www.askoxford.com/asktheexper...quency?view=uk
http://en.wikipedia.org/wiki/Letter_frequency

Here is a variation on Dave's Select Case with a slightly different syntax. Note: I'm not sure if (or when) there would be a tipping point between the two versions when more characters are added (ie either in a single line or as a separate Case statement).

Please Login or Register  to view this content.

-----------------------
From a different angle, I wonder if this could be solved using regular Expressions?
Unfortunately, I've only started reading up on these in the past few days & have never used Reg Ex myself so I can't offer a possible Reg Ex solution.

Shg, nice link. I'm not yet a real wiki fan but I couldn't help myself this time & succumbed to linking to it as well.

hth
Rob

I see what you're saying about loading an array into memory rather than reading it off the sheet each time Rob, thanks for the tip. I'll keep that in mind for any upcoming projects and try to use it so I get in the habit.

Administrative Note:

Hello sudeep and Welcome to Excel Forum.

We are happy to help, however whilst you feel your request is similar to this thread, experience has shown that things soon get confusing when answers refer to particular cells/ranges/sheets which are unique to your post and not relevant to the original.

Please see Forum Rule #4 about hijacking and start a new thread for your query.

If you are not familiar with how to start a new thread see the FAQ: How to start a new thread

Let us know if you have any questions.