Lack of a Good Editor

[dbooth-boston]

'Recently both Atom and Visual Studio Code have both come to the
mainstream. While there are many plugins that will do automatic code
completion for you in Javascript, Java, etc. There is none out there
that will do that for you in Turtle."
https://lists.w3.org/Archives/Public/semantic-web/2018Dec/0088.html

[william-vw]

Perhaps this could solve the up-front dataset authoring cost to an extent - an editor, loaded with known predicates and types, that supports a sort of auto-complete while authoring RDF?

[draggett]

Another approach would be to focus on easier editing tools at a higher level than the RDF core. A higher level framework would embody richer knowledge that editing tools could exploit to simplify authoring, e.g. expanding and collapsing level of detail at meaningful semantic boundaries, tabular controls for property/value sets, n-ary chunks and so forth.

[william-vw]

Higher-level – would you mean up to the ontology level or still RDF-specific (syntactic) features such as different views (e.g., tabular) on RDF data? Personally, it’s long bothered me that Protégé is essentially the only tool to work with OWL ontologies. Not that it’s a bad system but it looks a bit dated and there’s no alternative (aside from WebProtégé, which looks pretty cool but is again from the same people). From: Dave Raggett <notifications@github.com> Sent: December-10-18 10:35 AM To: w3c/EasierRDF <EasierRDF@noreply.github.com> Cc: William Van Woensel <william.van.woensel@gmail.com>; Comment <comment@noreply.github.com> Subject: Re: [w3c/EasierRDF] Lack of a Good Editor (#35) Another approach would be to focus on easier editing tools at a higher level than the RDF core. A higher level framework would embody richer knowledge that editing tools could exploit to simplify authoring, e.g. expanding and collapsing level of detail at meaningful semantic boundaries, tabular controls for property/value sets, n-ary chunks and so forth. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#35 (comment)> , or mute the thread <https://github.com/notifications/unsubscribe-auth/ALT1sBBpSgZmAhulbvHiQgwYlVlUv_K2ks5u3nEugaJpZM4ZKenE> .

[jeff-zucker]

Please take a look at https://github.com/jeff-zucker/solid-ide - it uses the Ace/Cloud9 editor to explore Solid data stores. I'd value feedback and suggestions to move it towards the kinds of things you are talking about here.

[ktk]

FYI we are experimenting with DSL frameworks for doing that. Currently I create R2RML mappings out of a DSL environment in Eclipse (Xtext). Benefit of DSL frameworks is that it takes care of a hell lot of stuff that would be a lot of work to implement on its own.

Most IDEs nowadays support the Language Server protocol protocol. Xtext can act as a LSP server so once we have a proper implementation there it should work on pretty much every IDE (given that you run the server as well or use it as a service). We experiment with creating Turtle as well in xtext with completion for vocabs & many other nice things.

Again the drawback is that you will need an xtext instance bu the benefit is that you implement the work once and will be able to use it in every editor that supports LSP. At least that is what LSP claims, we plan to see how well this works already. In my opinion this is the way to go for creating/maintaining RDF in some manual form. Which is surely not always the right way to go.

[jeff-zucker]

Thanks for the information. My project is less ambitious and probably no help to yours. I agree that manually maintaining RDF is not the way to go for most things. My target audience is those unfamiliar with RDF who just need to see what it looks like and see what happens to it when they make minor changes it. For that audience, things like auto-completion might be more confusing than helpful, depending on how they are implemented.

[draggett]

@darth-willy wrote

Higher-level – would you mean up to the ontology level or still RDF-specific (syntactic) features such as different views (e.g., tabular) on RDF data?

I have some ideas of my own, but really want to hear from industry developers as to what use cases, challenges and ideas would best help them. How to reach that far, though?

Current ontology tools are hard to work with - what would ease that? Perhaps better visualisation and browsing tools? Perhaps easier rule languages to work with given ontologies?

I've also been thinking about ideas for automatic induction and maintenance of ontologies from curated collections of examples and counter examples. Such a framework would also be designed to ease maintenance of rule bases to keep them in sync as the ontology evolves to match changing needs. In this sense, AI has plenty of potential to amplify human effort.

[draggett]

A related idea is to see how far we can go with using natural language for both data and rules. This assumes a graph representation that is close to the way words are used in natural language, rather than a deeper logical representation which is problematic given the looseness of natural language. It also runs counter to current practice of applying deep learning to natural language with no attempt to represent meaning except in a very limited and superficial way. One way to use natural language is as the input for machine learning of ontologies and rules. This can involve constraints on the grammar and scope for more robust processing, together with conventions for mapping natural language terms into those needed for the learned ontology.

[william-vw]

Yes, it would be highly useful to bootstrap the creation of ontologies and rules based on unstructured data (if that's what you meant). I quite appreciate Protege to work with ontologies (my main issue is that it's essentially the only ontology tool out there, afaik).

A solution could be, in analogy to the inference support, extend Protege with API support for creating ontology classes and rules, thus allowing third parties to create plugins.

For instance, to mine association rules, a solution could be to plugin the ML layer of the SANSA stack system (?)

It's true that, while Protege supports a nice (imo) editor for creating rules, it falls short when managing more than a few rules. I've done some simple work on improving its capacity for rule management - I was planning on generating a pull request after the semester.

[namedgraph]

Would anyone be willing to pay for a good editor or is open-source assumed here? :) We have the technology, it's just not where our focus is currently.

[william-vw]

Would anyone be willing to pay for a good editor or is open-source assumed here? :)

So either be prepared to pay for a good ontology editor, or move along (peasant)? I understand where it comes from but I don't think that's a very useful attitude.

We have the technology, it's just not where our focus is currently.

I'm sure you / we do, but operationalizing that technology is another issue. I am simply communicating a need that I (as a practitioner) have experienced, and my thoughts and efforts on filling that need.

Note that creating and maintaining ontologies is the only Semantic Web application that I've seen in university curricula (aside from niche courses on RDF) - such as knowledge management in health informatics, and data management in the school of information management.

[namedgraph]

What is a useful attitude then? (I'm not here to troll). My point is that commercial software comes with expectations like quality, maturity and support. Is that not part of operationalizing?
Would you expect to use a triplestore such as Stardog or Virtuoso for free as well?

Most of SW frameworks are currently open-source. Some are mature, such as RDF4J, Jena, librdf etc. But it took them 15-20 years to get there (!!!). Maybe the UX tools are just following the same timeline?

[cyocum]

Would anyone be willing to pay for a good editor or is open-source assumed here? :)

So either be prepared to pay for a good ontology editor, or move along (peasant)? I understand where it comes from but I don't think that's a very useful attitude.

When I wrote this, I did not have an ontology editor in mind. I wanted a general RDF editor which would have tab completes for predicates based on prefixes and things that my collaborators and I needed to create RDF data. Basically, my project we had create TRiG files which represent genealogies in early Irish. We have a few automated ways of creating parts of RDF/TRiG files but we need people to read and translate Old/Middle Irish into RDF/TRiG. So, I would love an editor (Eclipse, IntelliJ, Visual Basic Code, Atom, whatever) which can help them make the format correct and generally tell them when it is wrong and suggest predicates based on prefixes. Just like an IDE would normally do for programming languages.

We have the technology, it's just not where our focus is currently.

I'm sure you / we do, but operationalizing that technology is another issue. I am simply communicating a need that I (as a practitioner) have experienced, and my thoughts and efforts on filling that need.

Note that creating and maintaining ontologies is the only Semantic Web application that I've seen in university curricula (aside from niche courses on RDF) - such as knowledge management in health informatics, and data management in the school of information management.

I come from a Humanities/Software Engineering background and many Humanities data sets are human curated and maintained. From the small slice of the Digital Humanities that I am aware of, a good RDF data editor would be great and give people feedback faster on syntactic correctness, etc.

I also believe this would have wider benefits for people coming to RDF from other programming languages so they have same support for the RDF Serializations in their editors that they have for their programming languages. I have used Protege in the past and it is fine as far as I can tell for editing ontologies (Tbox) but it is not for editing data files (Abox).

[william-vw]

We're talking about ways to improve current (ontology) editing tools to suit practitioner needs - merely waiving your hand at that effort by saying "well, pay for it then" doesn't seem to contribute anything.

Also, who would I be paying? In lieu of such an alternative product (free or otherwise), would my only option be creating my own ontology editor? Even in that case, I would need to know what practitioners need.

Btw - Protege is at the tail end of the timeline you suggest (released close to 20 years ago). But even if it was a novel tool, wouldn't it be useful to look at ways for improving it?