Ideas

[sr258]

I've giving your idea some consideration and I think it's really interesting. I work as a secondary school teacher and automatically graded texts could be an excellent supplement to marking them myself, which I rarely come around to as often as I want.

Here are my thoughts:

It would be useful, if the teacher could also see the texts and the feedback given to the student by the content type. That way you could check a random sample to make sure students take their work seriously. I haven't figured out yet, whether the current H5P integration into e.g. Moodle allows the teacher to see the student's responses that are stored with xAPI or if they are only accessible to the student. If this is possible, it could be done that way.

I think grading language as a second category would also be a useful feature (I teach English). This could include things like:

• was the text written in a paragraph structure
• does it use conjunctions to connect paragraphs
• how elaborate is the word choice
• how complex are the sentences
• spell & grammar check

The content creator could define which of these criteria he want's to apply and could further specify options, like expected complexity of sentences, etc. If the student didn't do too well in one area, he could receive pre-defined tips on how to improve.

Of course, this grading language would have to rely heavily on external libraries, most of this is impossible to implement on your own.

Furthermore, I think you should design the content type in a way that's language agnostic, providing the content creator with a choice of the language he wants to use. Even without grading language, getting plural forms of keywords, etc. means applying linguistic rules. If you hard-code these rules to deeply into the content type, it will only be useful for grading English, or whatever language you use.

[otacke]

Hey Sebastian!

Thanks for your input! I had not expected someone to find this repository yet :-)

Yes, there should be an option to have a look at the results. I'll add that idea, although I'll have to figure out a good way. Using xAPI is easy, but H5P (for a good reason) leaves storing, statistics, etc. to specialized software.

Taking the language into account automatically is something that I have thought about, too, but that's definitely something for the (machine learning) future and/or a free library that offers grammar checking. However, computing a Flesch-Kincaid index or some other established indicator might be a low hanging fruit.

Let's just start with the very basic "fuzzy comparison" features that I created for "Fill in the blanks" that should at least cover things like a plural s easily.

Again, thanks for your ideas! I'll add them to the list this weekend.