Christopher:
Thanks for you thoughtful reply; it highlights important issues.
I think I could have been clearer about my questions, but I was not sure of the
context, and I'm not an expert in this area, so here are my "speculations":
I think the issue is pretty general, in that it applies to ontologies other
than GO....because....almost all ontologies define categories that are linked
together in a hierarchical (or DAG) structure.
So think of the (an) evolutionary "tree". The components (nodes) are linked
(by edges) into a tree structure. (a subset of a graph)
The names of each component either don't have type, or we don't care what the
type is because they are just names, and the type conveys no "relationship" (as
with the implicit relationships among integers). But just as you suggested
with regard to cities, the relationships among the nodes may sometimes be
useful for "inference", and "mechanically," all that means is navigating an
ontology as part of the search process.
Suppose for example that you have a database of organisms that includes a
genus-species for each organism, AND you have the (an) evolutionary tree at
hand built from all organism categories from "root" to "genus-species". Then
you can use the tree to help answer questions like "Show me all the mammals",
because you can traverse the ontological tree "down" from mammal to the lowest
level ("genus-species"?), and then list ANY organism from the database
annotated with ANY category in the ontology "below" mammal.
Now in the case of GO, things are a little different, but the ideas are
similar. I might for example want a list of genes that have been annotated as
being involved in any "biological process" that is a "part_of" (as indicated
using the "part_of" predicate) a specific process. That is, I want to find any
gene annotated as being part of any process that is one hop "below" the
specific/target process.
With respect to the "is_a" relationship within the "molecular function" portion
of GO, I might want to ask for genes annotated as being types of "enzyme
activator activity" (GO:0008047), so I might like to find genes annotated with
GO:0008047 AND any category "one hop below" 8047 to find genes that are known
to display more specific biochemical enzyme activity like "helicase activity"
(GO:0004388) and "ATPase activity" (GO:0016887). (I'm using a mutated version
of GO for this example, by the way; these relationships are not necessarily
accurate.)
This seems a bit too mechanical to call "inference", but as far as I can tell,
that's exactly what happens "under the covers" if you apply inference rules
along with an ontology to a database.
Actually, this seems to me to be quite a powerful capability that may help
distinguish semantic from relational approaches. It's not that you can't do
things like this with relational databases, but I think it's "harder". For
example, you can easily write a SparQL query to list all the n-level
descendents or ancestors of a node in a tree, if you know "n". And then you
can write another SparQL query to search for all nodes annotated with any
member in the list. If you want to find ALL "descendents" you actually need to
iterate an indeterminate number of times, which SparQL does not support so you
have to enter multiple SparQL commands, arrange for inference rules to be
applied, or build and navigate a transitive closure on the DAG, as far as I
know.
Well, regardless of this discussion I like your interface a LOT....I think it
REALLY simplifies the query process.
When I present semantic technologies to normal people, they don't seem too
excited about writing SparQL queries (well, ok, they recoil in horror :-), and
your interface will provide another, much easier and more palatable,
alternative.
so...thanks again for your work, and I hope you eventually find time to enhance
it to support at least the basic "hierarchy-relative" queries,
:michael grobe
> we currently do not extract Gene Ontology data from Wikipedia.
>
> We extract data for http://en.wikipedia.org/wiki/Template:Infobox_protein
> which doesn't use Gene Ontology. The template
> http://en.wikipedia.org/wiki/Template:GNF_Protein_box
> (and maybe others?) uses Gene Ontology, but we don't
> have a mapping from that template to our ontology (yet).
>
> Besides, most (if not all) of the IDs we already extract (ISBN, Protein
> Data
> Bank ID, etc.) are defined as strings, not as numbers, because with
> most of these, ranges and other numeric operations don't make sense.
> Would it actually make sense to search for genes annotated with a
> GO category "between" GO:0005488 and GO:0008047?
>
> Not sure what you mean by "descending from" and "2 hops from
> GO:0005488", but I guess the DBpedia extraction and/or the
> browser would have to use some kind of inference to enable such
> searches. There are other domains in which such inference would
> be useful (e.g., the browser doesn't know that someone who was
> born in Berlin was also born in Europe), so maybe one day we will
> implement it, but it's currently not high on our to-do list.
>
> Bye,
> Christopher
>
>
> On Thu, Oct 1, 2009 at 17:54, Grobe, David Michael <[email protected]>
> wrote:
> > Christopher:
> >
> > I suppose this is a different issue, but since you're "on the line", I'd
> like to know if the interface supports searches on ontology ranges, so
> that I could, for example, search for any gene annotated with any Gene
> Ontology category "descending from" GO:0005488. or for any gene annotated
> with a GO category "between" GO:0005488 and GO:0008047, or any gene within
> 2 hops from GO:0005488, yada, yada.
> >
> > Thanks,
> >
> > :Michael
> >
> >
> >
> >
> >> -----Original Message-----
> >> From: Jona Christopher Sahnwaldt [mailto:[email protected]]
> >> Sent: Monday, September 28, 2009 5:24 PM
> >> To: Richard Cyganiak; Grobe, David Michael
> >> Cc: [email protected]
> >> Subject: Re: [Dbpedia-discussion] can't find Goethe using new faceted
> >> browser.
> >>
> >> Michael, Richard,
> >>
> >> you're correct, currently only the top values of each facet are used
> >> for auto suggestion, and the selected value must match exactly.
> >> When I find the time, I will extend the browser to use inexact
> >> matching for fields like 'name'.
> >>
> >> You can use the text search box at the top of the page to search
> >> for "Johann", though:
> >>
> >> http://dbpedia.neofonie.de/browse/~:Johann/
> >>
> >> Johann Wolfgang von Goethe is the second result.
> >>
> >>
> >> Christopher
> >>
> >> On Mon, Sep 28, 2009 at 19:32, Richard Cyganiak <[email protected]>
> >> wrote:
> >> > On 28 Sep 2009, at 17:09, Grobe, David Michael wrote:
> >> >> But that's the strange part....Using the new browser, when I click
> >> >> "Person to filter by "Person", I get a list of predicates, including
> >> >> "name". Then when I enter "jo" in the "name" text field, a pull-
> >> >> down list of names appears, starting with "John", "John Davis", etc.
> >> >>
> >> >> If I enter "john r", I get a pull-down containing "John Russell(7)
> >> >> and John Reiley (6)", and if I actually enter "John Davis", I get a
> >> >> list with a single "John Davis (18)" entry, which I can select for
> >> >> the John Davis content...
> >> >>
> >> >> ....but "johann" gets me nothing...not even "Bach or Pachelbel".
> >> >
> >> > Ah, now I see what you mean.
> >> >
> >> > I think that the auto-completion in the facet fields just covers the
> >> > top N values for the facet. You find "John Russell" and "John Reiley"
> >> > because there are several entries with that value (that is, people
> >> > with that name -- 7 and 6, respectively). All the names that I can
> >> > find in the autocompletion dropdown, no matter what I enter, show at
> >> > least five people with that name, AFAICT. "Johann Wolfgang Goethe"
> >> > just occurs once, and so does not appear in the dropdown.
> >> >
> >> > I don't know if it's a bug. Sounds more like an intentional
> >> > restriction or limitation.
> >> >
> >> > It's still bad though, it would be much more useful if all facet
> >> > values were used for autocompletion.
> >> >
> >> > (I can't find Christopher's email address, so I'm CC'ing Christian
> >> > Becker, please forward it ;-)
> >> >
> >> > Best,
> >> > Richard
> >> >
> >> >
> >> >
> >> >
> >> >>
> >> >> Very odd; it's as if Goethe is not in the database, or maybe German
> >> >> names are not being processed (neither of which sound likely).
> >> >>
> >> >> :michael grobe
> >> >>
> >> >>
> >> >>
> >> >>> -----Original Message-----
> >> >>> From: Richard Cyganiak [mailto:[email protected]]
> >> >>> Sent: Sunday, September 27, 2009 6:27 AM
> >> >>> To: Grobe, David Michael
> >> >>> Cc: '[email protected]'
> >> >>> Subject: Re: [Dbpedia-discussion] can't find Goethe using new
> faceted
> >> >>> browser.
> >> >>>
> >> >>> Michael,
> >> >>>
> >> >>> On 26 Sep 2009, at 23:08, Grobe, David Michael wrote:
> >> >>>> I can't find Johann Wolfgang von Goethe using new faceted
> >> >>>> browser...starting with "Johan" in the Persons section, but I can
> >> >>>> find him by starting with birthdate...
> >> >>>>
> >> >>>> can someone tell me why?
> >> >>>
> >> >>> It seems you can find him by searching for "johann" or "goethe" in
> >> >>> the
> >> >>> persons section, but not searching for "johan", "jo" or "goet".
> >> >>>
> >> >>> I suppose the faceted browser does only exact matches on words, and
> >> >>> doesn't search for partial matches and doesn't correct
> misspellings.
> >> >>>
> >> >>> Best,
> >> >>> Richard
> >> >>>
> >> >>>
> >> >>>>
> >> >>>> thanks,
> >> >>>>
> >> >>>> michael grobe
> >> >>>>
> >> >>>>
> >> >>>>
> >> >>>> ------------------------------------------------------------------
> ---
> >> ---
> >> >>> ------
> >> >>>> Come build with us! The BlackBerry® Developer Conference in
> >> >>>> SF, CA
> >> >>>> is the only developer event you need to attend this year.
> Jumpstart
> >> >>>> your
> >> >>>> developing skills, take BlackBerry mobile applications to market
> and
> >> >>>> stay
> >> >>>> ahead of the curve. Join us from November 9-12, 2009. Register
> >> >>>> now!
> >> >>>> http://p.sf.net/sfu/devconf
> >> >>>> _______________________________________________
> >> >>>> Dbpedia-discussion mailing list
> >> >>>> [email protected]
> >> >>>> https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
> >> >>
> >> >
> >> >
> >> > ---------------------------------------------------------------------
> ---
> >> ------
> >> > Come build with us! The BlackBerry® Developer Conference in SF,
> CA
> >> > is the only developer event you need to attend this year. Jumpstart
> your
> >> > developing skills, take BlackBerry mobile applications to market and
> >> stay
> >> > ahead of the curve. Join us from November 9-12, 2009. Register
> >> now!
> >> > http://p.sf.net/sfu/devconf
> >> > _______________________________________________
> >> > Dbpedia-discussion mailing list
> >> > [email protected]
> >> > https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
> >> >
> >
------------------------------------------------------------------------------
Come build with us! The BlackBerry® Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay
ahead of the curve. Join us from November 9-12, 2009. Register now!
http://p.sf.net/sfu/devconf
_______________________________________________
Dbpedia-discussion mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
