Tutorial

[harisont]

I've started working on a tutorial and I'm not 100% sure how comprehensive we want it to be.

I think we definitely want:

• at least one simple "universal" examples (I was thinking adjectives modified by adverbs and/or verbs with auxiliaries)
• depending on the system CQP/Tree is configured for, either @kajad's Slovene-specific ones or some Swedish-specific ones, which I may source from Márton's work and similar (or nothing)
• links to the depsearch docs (already there) and Grew tutorial

But:

1. do we want the examples to be in 2+ query languages? (I'd say yes, but I'm happy to exclude CoNLL-U and possibly deptreepy)
2. do we also want an explanation of the various settings?
3. is there anything else I am missing?

If the answer to 2 and/or 3 is yes, I propose that the info page should link to a tutorial, rather than contain the whole tutorial (if we go for this option, maybe we could consider hosting it separately, e.g. as a GitHub wiki or as a blog post on the SBX blog, so that it can be updated without making a new release of the software). 1 also takes a lot of space, but can be kept at bay by allowing the user to select one query language for all examples.
Or if we're ok with a very long page, we should at least have quick links at the beginning of it.

[Niklas-Deworetzki]

You already made some adjustments to this issue in #30

But here are my thoughts:

1. I think having the examples in multiple query languages is good, then people see that it's not just one that they don't like.
2. I don't think we want an explanation of the settings in the tutorial. Maybe, at a later point, we want something for advanced users that explains the settings. But even then, I feel like we either get users who will understand it without much explanation or users who might not want to engage with these at all.

I don't think that language specific queries are necessary. Given that we now have UD-mode, it is easier to have "universal" queries that make use of dependency edges and the default UD annotations.

However, it would probably be nice to show that you can access annotation layers that are not part of UD. Which makes it not language-specific but at least system-specific. But maybe the solution here is to hard-link a query to run on Korp or the Slovene system.
The benefit here is not that the results are in a certain language, but to show that you can use all the annotation layers that are already present.

[kajad]

Apologies for being late to the thread, but I fully agree with your proposal to keep the tutorial language-independent, lightweight and as a separate page, so it can be linked to from all possible cqp-tree installations (e.g. Korp and clarin-si).

We might, however, add 2-3 executable language-specific examples to the installation-specific About page (e.g. one showing an example search on UD layers and one involving additional ones).

[harisont]

Based on this discussion and #30, I'd say we can keep the current structure (short about page with a configurable link).

For the CLARIN.SI installation, the link will lead to an internal page with Kaja's examples in Grew and depsearch (but not deptreepy). Ideally, following @Niklas-Deworetzki's suggestion, I'd like to add one more, one that uses annotations that have no equivalent whatsoever in UD. I'll try to come up with something with error tags (meant for the learner corpus), unless @kajad has a better suggestion. (edit: I had forgotten that there was a perfect example in our shared doc)

For Korp, I'll decide later, but I'm thinking examples in Grew and deptreepy (not depsearch), some of which also on our learner corpus.

I'll try to complete everything for the CLARIN.SI installation between today and tomorrow, but I don't know if I'll make it in time with the Korp examples (and config that we talked about) too. But maybe these aren't not a must for us to make a release of this package either... or what do you think @Niklas-Deworetzki?