User talk:Markus Krötzsch

From semanticweb.org

Jump to: navigation, search

I do not read this page on a regular basis.

If you have a message related to Semantic MediaWiki, use the mailinglist:

semediawiki-user@lists.sourceforge.net

For personal messages use my email:

mak@aifb.uni-karlsruhe.de

[edit] Spam Targets

The Spamming on index.php continues. It may help if you lock that site and block the spammer-IPs.

I know you don't really have time for checking the Spam Targets regulary, so I candidate for Admin-rights, cause I check all the article-changes here nearly daily and therefore could do that. MovGP0 18:29, 14 April 2006 (CEST)

[edit] Old messages

Heya!

Max suggested to me that I should change "is a" to "a;" He thought it was a lot simpler that way.

It is irrelevant for now. "Is a" looks better in the Infobox. All of these will be replaced by "Category:" soon.

It doesn't makes any difference - except a cosmetic one in the infobox below. Also you might want to use both. This can made in two ways:

Current Version:

[[Relation:a]]    [[is a::Relation:is a]]
[[Relation:is a]] [[is a::Relation:a]]

Or in the simpler (but future) version:

[[Relation:a]] [[is same as::Relation:is a]]

btw: I prefer the "is a" Version. And at least the statement would be more meaningfull if you would keep the "is" and throw the "a".

MovGP0 22:30, 14 November 2005 (CET)

Agreed. The problem with mixing up "is a" (in whatever wording) with "Category:" is that the system will probably not be able to reconcile the two. So users looking for cities might either look for any article which "is a city" or for all articles with "Category:City" but they will never get all results. This is why I would prefer to extend the existing category system instead of introducing an additional "is a". I also remark that we are not close to really supporting the "is same as" construct given above; sure you can write it this way, but it is not taken into account for answering queries and the like. --Markus Krötzsch 09:11, 15 November 2005 (CET)

[edit] Datatypes

Hi,
sorry to make this comment within your discussion page - but I'm looking for some developer-feedback about.
Built in Types
Thanks, MovGP0 18:22, 18 November 2005 (CET)

[edit] Regression bugs

In the recent software update, some functionality appears to be lost.

~~It seems semantics included by templates is not recognised anymore (see the yellow box on my user page) - I always regarded that feature as higly desirable.~~ PASS
Template trouble (URL encoding): (Relation:located in, has relation property, transitive)^{in&property=has relation property&value=transitive (+)}

used to display nicely...

--Joris Gillis 22:43, 22 February 2006 (CET)

Hi. I just noticed this message -- I added my email addresses above for future communication. Concerning your questions:

We did not remove template support, but it is not enabled by default (since it requires a MediaWiki patch). After upgrade, just go to SMW_Settings.php and look for "template" to find out what to do.
The second thing is queer, but so is the nested use of templates. Is it actually expected that MediaWiki supports recursive template inclusion?

--Markus Krötzsch 11:26, 2 March 2006 (CET)

[edit] Thanks

... for moving my page! --Langec (aka "Christophe") 10:49, 15 July 2006 (CEST)

[edit] Spam test

...

[edit] Bug

Thanks for making me sysop.

I have looked at the mysterious bug 1, which behaves in a systematic way, but I still do not know how to fix it, as e.g. deleting and recreating pages does not seem to work.--Patrick 01:04, 7 September 2006 (CEST)

Yes, I also find this bug quite strange, but your narrowing down of the affected articles will help further investigation. I have a local mirror of ontoworld.org in which I can do some further playing around. The tables for the semantic data are correct for the affected articles, so I assume that the problem is caused by the current code for resolving redirects. If it is systematic, it is strange that this bug affects only a handful of pages, and that it was not reported by any other site using SMW.

Also note that some further hints can be obtained by using the optin "debug" with the problematic ask queries. An example is given below. It clearly shows that quite a number of unrelated articles are pulled into the query. --Markus Krötzsch 11:27, 7 September 2006 (CEST)

AWIC2007, CS-TIW2008, France (European part), French Guiana, Guadeloupe, ICCS2008, ICFCA2007, Jurix 2006, Martinique, Nancy, OTM2006, Paris, Réunion, SAC2006, Secureware 2008, TABLEAUX2007, TWOMDE 2008, WISE2007, and XTech2007

I see, so the database is fine, the problem is the program that translates an ask-query into a wrong SELECT query.--Patrick 15:52, 7 September 2006 (CEST)

17 of the 40 problem pages are reported in http://wiki.ontoworld.org/index.php?title=Special:Whatlinkshere/Help:Relation&limit=500&from=0 to redirect to Help:Relation, although they don't.--Patrick 17:31, 7 September 2006 (CEST)

Wikible has the same bug, with other affected pages, also apparently related to redirecting, see http://wikible.org/en/User:Patrick#Bug_1 --Patrick 17:42, 7 September 2006 (CEST)

Ok. the only current workaround I can offer is to switch off redirect-resolution by setting

$smwgIQRedirectNormalization = false;

in your SMW_LocalSettings.php. Of course, this disables redirecting completely. --Markus Krötzsch 19:53, 7 September 2006 (CEST)

Okay, let us do that on this wiki, until the bug has been fixed. I think it is more important that the queries give correct results.--Patrick 20:08, 7 September 2006 (CEST)

The existing redirects are still useful as soft redirects.--Patrick 20:20, 7 September 2006 (CEST)

I have removed the redirects (partly deleted, partly changed to "see ..". That seems to help.--Patrick 16:08, 9 September 2006 (CEST)

What I thought was the same bug on Wikible may be less serious: although in-page queries asking for pages with a given relation to enumerated pages only work if these enumerated pages are in the main namespace, a redirect from Glossary B to Glossary:B causes that a relation to B counts as a relation to Glossary B.--Patrick 16:49, 9 September 2006 (CEST)

Sorry for not replying lately and thanks for doing the fixes. Unfortunatelly, I am very busy these weeks. We are preparing to collect information about Semantic Web tools, vocabularies, and languages with this wiki, and a news item will appear on the front page soon. Otherwise I did not get much work done on SMW recently. Maybe next weekend. --Markus Krötzsch 15:00, 13 September 2006 (CEST)

[edit] Attribute:Located in European Union etc.

Hi,

You marked templates like Attribute:Located in European Union as deprecated. However, they are used in tables like that in Europe, to present data in a compact and convenient way that is not possible with relations and categories. Note also that the data are not quite binary, sometimes "partly" is used, and sometimes we want to distinguish between "no" and "not yet determined/annotated". See also Help:Category#Attribute.

Patrick 01:27, 5 November 2006 (CET)

I do not fully agree. Of course, the use of multiple string values for an attribute is possible, but a value "v" is not self-explicatory. If a property for this purpose is really needed, then several things must be fixed:

The attribute must be documented, especially its possible values need explanation.
The attribute must get a better name, since it currently misleads the reader to think it is a Boolean for [[located in::European Union]]. What you want to express is not related to phyiscal location, but to legal membership status. A possible name would be "European Union membership" or something similar. This makes clear that one needs to read the docu to understand the intention.
The possible values must become self-explicatory. So essentially "v" needs to be replaced by something like "full member".

I added explanations at attribute pages and at some tables using these marks/codes. In general it is very common in the display of information (tables, maps, etc.) to use an abbreviation, code, or symbol to save space, accompanied by a legend.--Patrick 23:59, 5 November 2006 (CET)

Abbreviations are usually used in cases where there is not enough space (e.g. on a map). But people today have large screen and we have almost a terrabyte of free disks on the ontoworld-server, so there is no need for being economic rather than readable. People who come to an article should be able to understand the content by reading it, and without looking up some legend elsewhere.--Markus Krötzsch 13:45, 6 November 2006 (CET)

What I still don't understand is why, instead of completing ontoworld.org in cases where we currently lack content (e.g. we still don't have all countries), we choose to introduce more and more annotations, and especially those that are not easily expressed with SMW. Can't we just finish the use of the obvious annotations first?

Also, ontoworld.org is not a wiki about countries or places (or movies for that respect!). We only enter countries and towns since they are relevant as locations of events (and since they are handy for demos). So we should not continue to add all kind of information to ontoworld, but just those kinds which are clearly on-topic (and there is so much to do there as well).

Pages not about the Semantic Web are intended for trying out and demonstrating the possibilities and limitations of SMW.--Patrick 00:10, 6 November 2006 (CET)

I know, but features like RDF export, recent changes, etc. still bring people's attention to such content. If someone ends up on a very experimental page, this will give a strange impression of the wiki in general. I also think that much of the current testing on ontoworld.org does not help much, since it demonstrates how one should not use SMW rather than how to use it. We already know the features and limitations of SMW quite well, and we should try to make ontoworld.org a good example of how to use SMW properly. For example the Template:ask is not very nice anymore, since it executes numerous queries that are not needed at all, hence wasting performance. It also creates output that is somewhat unpredictable, and which often is not is the shape for being inserted into articles. I think the idea of Template:ask was to have an instant query for atttribute and relation pages, but not for all other kinds of articles as well. Currently, ask is used on many non-testing articles. So: yes, we should do some testing, but we should be more concerned with the proper content (people, events, papers, semantic software tools, ontologies/vocabularies, stubs for countries, cities, software licenses) than with test pages. Real content is the best test for SMW. --Markus Krötzsch 13:45, 6 November 2006 (CET)

I think it would be a good idea to have some basic attributes and relations worked out before the eventual addition of the Semantic MediaWiki extension to Wikipedia. Fuelbottle 01:02, 6 November 2006 (CET)

Yes, I fully agree, though I think that ontoworld.org is somewhat extreme in its attractivity for creating strange and experimental attributes etc. But it would be useful to have some basic relations/attributes/categories as well as more specific annotations for one given domain (e.g. countries and geography). One could then quickly annotate the chosen domain as a gold standard for proper annotation. --Markus Krötzsch 13:45, 6 November 2006 (CET)

--Markus Krötzsch 03:00, 5 November 2006 (CET)

P.S.: Edits in the Help: namespace, unless they are minor fixes, shoud be discussed with the developers, since the Help: namespace of ontoworld.org is considered by many as the official docu for SMW (even if a page is not listed in the docu). Neither Denny nor I usually do large additions there without discussing it first.

I see. I thought it was like MetaWikipedia and many other wikis, where users can edit in the Help namespace just like in other namespaces.--Patrick 00:14, 6 November 2006 (CET)

Yes, and we also like people to contribute documentation there, but it should be documentation that all of us can agree upon. For instance, someone could easily create a docu on how to run SMW with another database or on another system. Or one could have an FAQ that provides technical help that was found to be relevant on the mailing lists. Such documentation is undiscuted. But Help:Category#Attribute gives hints on using annotations with agree. So this page already presents things as solutions before they have been discussed with anybody else. What we need is more discussion to coordinate the progress of ontoworld.org.

I have adapted that section so that it recommends the use of a full word or phrase instead of a simplified checkmark.--Patrick 02:03, 10 November 2006 (CET)

Maybe we could meet sometime soon for this purpose? Face to face discussion is usually more productive than writing talk pages ;-) You could send me an email where you are currently located, and we can consider how to arrange a meeting. --Markus Krötzsch 13:45, 6 November 2006 (CET)

[edit] Answer to your email

Sorry for the delay in answering your email. I copy here selected parts with answers for convenient linking and for the case that others want to join the discussion.--Patrick 17:05, 6 November 2006 (CET)

No problem. Looking at this page, I see that MediaWiki si really not a great tool for having discussions ;-) We should rather continue this via mail ... --Markus Krötzsch 03:12, 7 November 2006 (CET)

Have you considered setting up an IRC channel on freenet? There is already one for MediaWiki, so why not one for Semantic MediaWiki? - Fuelbottle 17:59, 7 November 2006 (CET)

Not really. I would be happy if I had enough ime to hang around the relevant IRC channels that exist already. --Markus Krötzsch 15:41, 9 November 2006 (CET)

Some articles have become very long due to lengthy lists of <ask>ed queries that are inserted by {{ask}}. Since we mostly use ontoworld.org to show it to interested users, we would like to have a healthy ratio of text content and semantic stuff. If an article has only three lines of content, this should not be followed by three pages of query tables. The information given by the queries often is not very relevant either.

I just came across a result table labeled "location of location of page" and many other strange tables that I did not understand.

"Location of" is used for the next higher level, e.g. the location of Eindhoven is North Brabant, so the "location of the location of Eindhoven" is the Netherlands. It is very logical, but admittedly not how one would express it normally.--Patrick 17:00, 6 November 2006 (CET)

We can rephrase it to "second higher location level".--Patrick 17:03, 6 November 2006 (CET)

Perhaps we should use CSS to make (part of) the result of {{ask}} invisible by default, while allowing interested users to see it. What do you think?--Patrick 17:00, 6 November 2006 (CET)

Rather than introducing more and more annotations, we should try to complete our use of the existing annotations. We still have not all countries, but some already have annotations for their official language.

That is normal for a wiki where the editors are volunteers.--Patrick 17:00, 6 November 2006 (CET)

Usually those annotations seem to be written down without caring much about the appearance of the articles. So some articles look like lists of annotation-worthy content, but have only very few real sentences. Other articles have long copied passages from Wikipedia which have many dead links.

That is partly because the annotations and queries are the extras compared with Wikipedia, so people focus on that. Partly it is also because, oddly, there are so few editors.--Patrick 17:00, 6 November 2006 (CET)

There are also articles where annotations are hidden by using empty alt texts (e.g. Africa).

The annotation appears in the factbox. Sometimes it seems superfluous to see them twice.--Patrick 17:00, 6 November 2006 (CET)

[edit] Automated cleanup

Moved from previous section, from your email: "Maybe Denny will do some automated cleanup and importing of further data to prepare the demos."

If the automated "cleanup" means large-scale undoing of people's work, please discuss it first.--Patrick 17:00, 6 November 2006 (CET)

Fully agree with that! Any large scale loss of work should be discussed. (Sorry btw, I did not update myself on this discussion fully yet) --Denny 17:20, 6 November 2006 (CET)

Same with me. The idea was not to delete articels or something like this. But one could use a script for removing deprecated annotations (e.g. Relation:is_a) from non-testing articles (such as countries and cities). But we mainly did new imports, which now add to the content of the wiki. --Markus Krötzsch 03:12, 7 November 2006 (CET)

In most cases, instead of just deleting deprecated annotations, they can better be replaced by more suitable ones. For example:

annotations of Relation:Instance of can be replaced by category tags, except that for countries these are not needed, because the country pages concerned are already in a subcategory of Category:Country.
for some attributes using "v" as checkmark, this "v" can be replaced by a word or phrase. For example, for Attribute:Located in Europe it can be replaced by "fully".

For annotations of Relation:Is a I do not care that much if they are just deleted, or replaced by category tags, it is just some old miscellaneous stuff.--Patrick 09:38, 7 November 2006 (CET)

Annotations of Relation:Is an instance of can simply be deleted. They are only on university pages which already are in Category:University.--Patrick 10:29, 7 November 2006 (CET)

[edit] Hallo

Hallo Markus, ich habe mich vor ein paar Sekunden angemeldet aber ich bin ein wenig verloren, da ich vor allem bei wikipedia tätig bin. Kannst Du mir zufälligerweise sagen, ob es eine Seite gibt, wo man Fragen stellen kann (eine Art Forum oder so was)? Und wo gibt es denn eine Seite die irgendwie erklärt, was das semantische wiki ist... für Leute, die keine Ahnung in Informatik haben (wie ich ;-) ). Ich danke Dir im voraus für Deine Hilfe. Grüße, Big Brother 18:07, 9 November 2006 (CET) ps: ich habe an Dich gedacht einfach, weil ich Deinen Namen vor einer Sekunden in den recent changes gesehen habe. Habt Ihr ein IRC hier? Big Brother 18:07, 9 November 2006 (CET)

Hi. Dieses Wiki ist eine Installation von Semantic MediaWiki und wird direkt von uns (den Entwicklern dieser Extension) betrieben. Semantische Wikis haben das Ziel, die Organisation von Informationen im Wiki zu vereinfachen und dadurch die Qualität des Wikis zu erhöhen. Da ich gerade auf der ISWC2006 bin, kann ich keine ausführlichere Beschreibung geben, aber vielleicht hilft ja Semantic MediaWiki und Help:Semantics. Das gesamte Wiki ist wie du siehst englischsprachig. Ansonsten sind wir am besten via Email zu erreichen, oder aber über die Mailinglisten zum Thema Semantic MediaWiki (siehe SMW). --Markus Krötzsch 18:16, 9 November 2006 (CET)

Wenn die Links dir nicht helfen, dann freuen wir uns auch über Hinweise zu eventuellen Unklarheiten oder möglichen Verbesserungen in der Doku. Dies ist ein relati kleines Wiki und hat deshalb nicht so eine umfangreiche Gemeinschaft von Editoren, die Fragen im Wiki beantworten. Email ist wirklich einfacher. --Markus Krötzsch 18:19, 9 November 2006 (CET)

[edit] Articles about classes

If articles about classes e.g. Category:Person will be in the category namespace rather than the main namespace, perhaps a show/hide button similar to what is used on infoboxes on wikipedia (e.g. at bottom of http://en.wikipedia.org/wiki/Germany) would be needed to be able to hide the list of articles and subcategories of a category? - Fuelbottle 21:07, 9 November 2006 (CET)

I would not say that Category:Person is really the article about the concept "person". It is a category article which should describe when to use the category, and what else is relevant when annotating person. So it is more like a editor documentation than like a real article. Having hide buttons is always handy, but do you think the lists of articles and subcategories really are in the way? After all, they are below the article anyway. --Markus Krötzsch 07:23, 11 November 2006 (CET)

After you recommended avoiding Relation:Instance of and using the category system instead, it seems convenient to use just the category page for a class, avoiding the duplication of having pages in two namespaces for each class. The inconvenience of having the factbox in the middle of the page is solved with __TOC__ and a header ==Factbox==. However, the fact that a category cannot be renamed is a drawback. See also Help_talk:Category#Category:Automobile_etc.--Patrick 10:43, 11 November 2006 (CET)

I think this is an issue that each wiki community can decide from case to case. In general, the technical/navigational info contained on pages like Category:Person is quite different from what you would write on wikipedia:Person. For some wikis, it might be convenient to combine this into one article, but it is probably not a great approach in general since wikipedia:Person just is not concerned with question like what FOAF URI you use in RDF feeds. You want to protect your readers from too much technical details that are closer to an editing guide, as you find on category pages. --Markus Krötzsch 21:29, 11 November 2006 (CET)

The problem is that exported data should be consistent and clearly distinguish classes from instances. This could be done either by the export script or by processing the data after it's been exported, both requires classes and instances to be marked in a consistent way. One possibility would be to put everything that's a class exclusively in the category namespace, on wikipedia this would be many things like broccoli, car and shampoo. Another possibility is to use Relation:Category about to ensure that ensure that external software would be able to use both the class hierarchy made by the categories and attributes and relations found in the corresponding articles. An article could also be marked as a class by adding Relation:Has subclass to the parent category, for example Category:Brassica would have Has subclass::Broccoli. Fuelbottle 19:34, 13 November 2006 (CET)

The link between classes and instances can be established informally via the name of the article. It would still be OK to write software that takes this into account for some user interface adjustments (e.g. to show corresponding categories for displayed instances, if available). To make categories acceptable as classes, one needs to interpret them as "the class of all things which have an article in category X" and nothing more. Especially, classes do not always imply "is a" relationships. This is not a problem, since it still makes sense to treat categories as classes (i.e. the way classes are handled in tools is mostly adequate for categories). Future semweb languages might have better mechanisms for making links between instances and classes explicit (e.g. via "punning" in OWL1.1). A problem might be in interpreting the subcategory relationship as subclass relationship, since the category "tree" in Wikipedia is rather messed up in this respect. Maybe this will improve now that the category browser extension is there, or we could provide the explicit "subclass of" special property. --Markus Krötzsch 19:35, 16 November 2006 (CET)

[edit] Reification

What do you think of making statements reifiable to have the option of adding additional information, see User talk:Patrick/annotation ? - Fuelbottle 21:07, 9 November 2006 (CET)

Thanks for the note. See my reply on Patrick's talk page. --Markus Krötzsch 07:23, 11 November 2006 (CET)

[edit] n-ary relations

BOWiki is a fork of Semantic MediaWiki that includes n-ary relations. Are there any plans for integrating this into the official Semantic MediaWiki? It would seem this might be a solution to the problem that it's currently not possible to define things like date ranges for relations such as "president of". - Fuelbottle 00:08, 20 November 2006 (CET)

Interesting, but the syntax becomes quite complex for occassional readers, doesn't it? There are no plans to enable n-ary relations in SMW, which does not mean that we think that they should not be possible -- we just don't have resources for this at the moment.

The president example you mention is an interesting special case. Wouldn't it be a good idea to have special simplified support for associating times (start+end, or just start) with relations? One could, e.g., have a syntax like [[president of::USA (January 20, 1961 – November 22, 1963)]] instead of writing [[president of::country=USA; start date=January 20, 1961; end date=November 22, 1963]]. The special case of times would cover some of the most important cases for n-ary relations without making the syntax overly complex. Of course, one would have to check the details (do we want this for every annotation (then we should probably use [...] to avoid confusion with valid article titles, or should one have to declare the timedness explicitly with each annotation? Another challenge is to implement this somewhat efficiently, and to make it compatible with untimed versions (so if I just write USA this should still work and be askable). Much work but probably worth it (especially now we have Timelines :-). --Markus Krötzsch 12:16, 20 November 2006 (CET)

The syntax might be a little complex, but with additional simplified syntax for associating times I don't think it would become too messsy. I think perhaps something like [[president of::USA {January 20, 1961 – November 22, 1963}]] would be a good syntax to avoid confusion. The software might then check for {} in an annotation to see if it was timed or not.

Have you looked at the code for BOWiki? Could their way of doing n-ary relations be integrated into the current version of Semantic MediaWiki, and extended with a simplified syntax for time annotation?

I'm very interested in seing this functionality so perhaps I could help with this. Fuelbottle 20:10, 21 November 2006 (CET)

[edit] Pages with problems

Thanks, the other pages are:

Patrick 19:40, 2 December 2006 (CET)

OK. It might take me a moment to fix this, since I did not manage to reproduce the problem on other servers so far. --Markus Krötzsch 12:34, 4 December 2006 (CET)

Thanks. I recovered the pages.--Patrick 17:41, 4 December 2006 (CET)

[edit] Spam Bots

Hi,
I've written a new Spam blacklist, maybe you might want to review it. — MovGP0 16:32, 29 April 2007 (CEST)

I fear that those keywords and URLs are likely to change quickly. But lets hope that the simple captcha extension suffices to repell the bad guys for now. --Markus Krötzsch 18:49, 29 April 2007 (CEST)

[edit] Typo in news on Main Page

Hi,

I could see that the first news item on Main Page contains date as 'July 5, 2005'.

I guess it can be 'July 5th, 2007'.

Thanks.

With regards,

Harshal 07:53, 27 July 2007 (CEST)