Apache OpenOffice (AOO) Bugzilla – Issue 74198
spell checker proposes to separate words which he don't knows
Last modified: 2014-02-24 17:29:38 UTC
The german words "drittstärkstes" and "sozialräumliches" for example are not known by the spellchecker (The german language makes largly use of composita, and much of them the spellchecker doesn't know) and are marked as wrong. So far no problem. But the spellchecker suggests "dritt stärkstes" and "sozial räumliches" - and both are definativly wrong. In german it isn't always clear if an expression is written as one or as two words, but the bad suggestion brings even much more confusion. When the suggestion is composed of two words, it is in german almost always a bad suggestion. So combinations of two words should never be suggested when checking german texts. I asume that for languages who makes large use of composita the situation will be similar.
Thanks for the report. I'm adding Bjoern Jacke (maccy at ooo), author of the German OOo dictionary to this issue.
this special cases of compound words of "ordinal+superlative" are hard to put into a compound *rule* and on the other hand the number of this kind of compound constructs is very high but while most of the possible compounds are rarely used. As long as I don't see reasonable rule to support this class of compounds, I think it's best to add the most common ones that are not already known manually. Laci, you can assign this to me if you want...
Okay. What's about the idea to no longer suggest strings of two words? There are some cases where this makes sense (old orthographie: fahrradfahren, new orthographie: Fahrrad fahren). However, in practice I've made the experience that most of the suggestions containing of two words are bad.
MySpell/Hunspell has an affix file option (NOSPLITSUGS) to forbid this sort of suggestion. It is implemented for Dutch, because Dutch dictionary lacks of compound word support, and most of its suggestion was wrong. But it is not true for Hungarian and German. Word-level spell checker of OOo suggests two "good" solution: add "drittstärkstes" and "sozialräumliches" to the custom dictionary or split these words. Naturally the best solution is to add these words to the German dictionary, like most of the other compound words of German.
I often see these suggestions and I also think that that' annoying, but on the other hand, from time to time I "forget" a blank and am thankful that spellchecker does not accept my sentence "Ich fielhin und schlug mir ein Knie auf", but suggests "fiel hin" for "fielhin". So we should not overshoot the mark, it's correct that the spell checker suspects that the user might have created accidentally a compound of 2 words. An option to forbid those suggestions might be a good solution for users who do not have that "missing blank problem". I do not see this as a DEFECT, but as an ENHANCEMENT.
>> I often see these suggestions and I also think that that' annoying, but >> on the other hand, from time to time I "forget" a blank and am thankful >> that spellchecker does not accept my sentence "Ich fielhin und schlug >> mir ein Knie auf", but suggests "fiel hin" for "fielhin". It may be nice to have this suggestion. However, when the word "fielhin" is underliened, it is clear to everyone where's the problem. Also without suggestion. On the other hand, "sozial räumlich" is a _wrong_ suggestion for "sozialräumlich". Also when this word is attached to the list, there will stay other words with wrong suggestions. And that makes a very bad impression to the user: When I see a long word which is underlined, I normaly leave it how it is because in most cases it is a problem of the spellchecker. I've stopped even to use the suggestions because many of them are bad. And I don't think I'm the only user doing so. This price to high in comparation with the very small benefit.