The previous version of LanguageTool we used was 1.8. Now, we upgraded to release 2.2. Since then, some of our Unit tests fail because the sentence begin error detection doesn't work properly any more. Depending of the entire text to be checked, in some cases LanguageTool doesn't detect if a sentence starts with a lowercase letter.
etwas beginnen. und der auch nicht
Result: Only "etwas" will be marked as sentence begin error, but not "und"
Enter following text:
etwas beginnen. und der auch nicht. weil dann gehts fast
Result: Now, "etwas" and "und" is marked as wrong, but "weil" is missing
Switch language to English
this is a sentence. and this as well.
Result: "this" is marked as wrong, while "and" is not
Enter following text:
this is a sentence. and this as well. what is going on
Result: "this" and "and" are marked, but "what" is missing
Enter following text:
this is a sentence. and this as well. what is going on?
This is a fatal issue in our environment because right now, the grammar check appearence is inconsistent.
One further strange behavior:
I have taken a look at the code. When you end a sentence properly with a punctuation mark (.?!…), everything is OK. This behavior was introduced (in some languages) in order to avoid false alarms in tables (which can contain sentences starting in lower case and without a punctuation mark at the end). Any solution has pros and cons, as there is no way to know if a sentence has to end with a punctuation mark (titles, tables, etc., usually do not).
Hi Jaume.
Thanks for you analysis!
If we are within a table cell and somebody writes following:
Then he forgot to set the punctuation mark at the end and there's also a sentence begin error ("und"). In my environment, no errors would be indicated by grammar checker. Our customers wouldn't agree to this behavior.
Would it be imaginable to make this behavior configurable?
Hi Andreas,
I think you are right in your claim. The errors in your examples should be definitely detected. But I think that to make this behavior configurable is not a good solution. We'll try to find which is the "least bad" solution.
Hallo Jaume.
Thank you very much! Your Fix solved my problem! Good work! :-)
Cheers
PAX
Last edit: Andreas PAX Lück 2013-07-03
Well. This is not definitive. The current solution causes a lot of false alarms in Wikipedia checks. I will try a midway solution, and I hope we can reach an agreement among the mantainers.
I get some false positives because the sentences are not split correctly. For this German text:
I get a warning that "Zur" should be lowercase, because it is not at the beginning of a sentence. I don`t get this error when using the spellchecker on the www.languagetool.org. But when i use the fallback form (http://www.languagetool.org/simple-check/) i can reproduce this problem.
Is there any way i can workaround this problem? Or is there any reason why the spellchecker on the homepage works, but not on the fallback form? (I`m using LanguageTool 2.2 from maven central)
Thanks
Emmeran, I cannot reproduce either of your issues. Which browser are you using? Which language did you select? If spell checking doesn't work that's often caused because a language without a variant ("German") has been selected instead of e.g. "German (Germany)". I suggest you open a new issue (or even better, post to the forum or mailing list) so this one doesn't get too crowded.
Jaume, can this issue be closed?
Daniel,
I think it can be closed. The current solution is not perfect, but it is reasonably good.