Posted by taps128 on 11/28/07 12:58

Rik Wasmus wrote:
> On Wed, 28 Nov 2007 10:44:14 +0100, taps128 <nstjelja@gmail.com> wrote:
>
>> Hi, all. I have a poroblem, and I'd really appreciate if someone
>> helped me to solve it.
>>
>> The problem is, I want to take the first two sentences of a string. To
>> do that i need to split them whenever a dot occurs, and the join the
>> first two array occurences in a new string but I have a problem
>> beacuse the dot in the Croatian languages is not always used a
>> sentence delimiter, but is often used in conjuction with numbers and
>> acronyms. So I wanted to use a regular expression to split a string on
>> every dot ocurence but not when a dot is precedeed by a number or a
>> 'd' or a 'o'.
>> This is my best shot at it:
>>
>> $string='Glavna skupština Društva će se održati 27.12.2007. (četvrtak)
>> u 11 sati u prostorijama Doma hrvatske voj­ske u Lori u Splitu.Atlas
>> turistička agencija d.d. stekla je 22. i 23. studenog 2007. godine
>> 2800 vlastitih dionica.';
>> $uvod=preg_split('/((d\.o\.o\.)!|(d\.d\.)!|[0-9]!)|\./', $string);
>> print_r($uvod);
>>
>> But it doesn't work right. If someone knows how to slove this problem.
>> Any help will be really appreciated.
>
> Well, you main problem here is to decide WHEN a dot is ending a
> sentence. Not a very simple task without lists of known acronyms. Also,
> a dot after a number can end a sentence:"The coldest winter I remember
> was in 1985. The temperature the day my sister was born dropped as low
> as -21C.". How do you propose this is handled? Formulating the exact
> requirements before writing the regex is more then half the work.
>
> A start for you (by no means complete):
> A sentence is ended by a dot:
> - followed by either $ (for which case we don't need a regex, it will
> end automatically there), or:
> - at least one whitespace character (\s) (well, it should be, damn those
> kids nowadays), followed by a capital letter (\p{Lu}, use utf-8 mode).
>
> That would be:
> preg_split('/\.\s+(?=\p{Lu})/u',$string);
> ...which doesn't split your string anywhere, as my rules for
> 'sentence-ending' seem to be inadequeate for your string, or no sentence
> is ended.
>
> I think this will require some hefty '(not) pre/proceded by' operators,
> as you'd like the matched text to be in the split. Even then a 100%
> success rate will most definitly by out of the question.
> --Rik Wasmus
tnx, that was what i feared.

[Back to original message]