Difference between revisions of "Classroom:Norwegian Grammar Checking"

Revision as of 17:49, 26 June 2011

Norwegian Grammar Checking

On clicking on the icon below (i.e., the picture of a troll), you will come to an automatic parser for Norwegian designed to provide grammaticality feedback on sentences of your choice.

Please click on the icon. In the window that comes up, feed a sentence into the parser by writing or copying it into the open mask, and push the button 'Analyze'.

If the string you write in is grammatically wellformed, you will get the response:

          A grammatically wellformed sentence of Norwegian

However, if the string is ungrammatical, the system will notice, and inform you in what respect the string is in violation of Norwegian grammar. For instance, consider the ungrammatical string (the '*' in front signals that it is ungrammatical):

          * Jeg liker du.

Here, "du" is used as an object, while this form can only be used as a subject. If you send the sentence through the parser, this is also what the parser will tell you - you get the following message:

      The word "du" is marked with the wrong case, try using "deg" instead. 

In the interface of this message, to the left below this message, there is a small square with a number inside. Click on this square, and you get a new window showing, as 'generated sentence', the recommended version:

         Jeg liker deg.

To write a new sentence, return to the window prior to the 'generate' window.

You can in principle use sentences of any length, but for the purpose of getting rapid feedback and precise advice about what (if anything) is wrong, we recommend using short sentences.

Using the grammar checker in synchrony with a course in Norwegian

The parser can be used in synchrony with the course NoW (http://www.ntnu.edu/now) given at NTNU, starting in Fall 2011. The parser uses the vocabulary of the course, and in various ways ties in with exercises used there.

For instance, for a text like one for Lesson 5 in NoW, called "Maria", illustrating inflections for Past tense and recast on Preterite/Past tense, for a prompt question such as

Write 4-5 sentences in answer to the question: "Hva gjorde du og Anna?"

you might write the following:

Anna og jeg syklet til en kafe. Hun er min naboen. Vi snakte i to timer. Ingen stjelte syklen min, selv om den ikke var låst.

If you write the sentences first in a normal text editor, spell-checkers in the editor may detect wrong spellings, which is not a primary task of the grammar checker. You may use those spell-checkers as a first sift, and feed the corrected sentence strings into the parser.

Relative to each phenomenon for which the parser gives an error message, TypeCraft will have a page with information about the phenomenon involved. Such a page will include sample error messages, short definitions, examples of correct forms for the phenomenon, and further excercises designed. More detailed explanations and reference to literature, as well as reference to other languages manifesting the same or similar phenomena, will be provided in the future.

Phenomena for which you can get feedback

Apart from the the feedback that the input string is a grammatical sentence, and that a specific grammatical error has been found, the parser can als inform you that a given lexical item is not in the lexicon used. Moreover, there will be cases where the grammar recognizes a string as ungrammatical, but no specific message has been designed yet - in such case the interface says "No mesage".

Occasionally it may also happen that the parser gives no response at all, which can be for a number of reasons of processing factors. Most likely, however, in such a case the string is not being recognized as grammatical.

For an overview of phenomena that you can get articulated error feedback on, please go to the following page: Phenomena