You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@stanbol.apache.org by Luigi Selmi <lu...@gmail.com> on 2013/02/19 10:45:43 UTC
chains/enhancers configuration
Hello,
it seems that all enhancements are stored in a single triple collection
org.apache.contenthub.enhancements. It will be useful to add a property to
chains/enhancers to be set at run time from the felix Console to address
different triple collections. Does it make sense ?
Best
Luigi
Re: chains/enhancers configuration
Posted by Rupert Westenthaler <ru...@gmail.com>.
Hi Suat
On Wed, Feb 20, 2013 at 8:41 AM, Suat Gonul <su...@gmail.com> wrote:
> Hi Rupert,
>
> LDPath filters are not considered in the Contenthub for the time being.
> As far as I understand, when you have a set of entities, and execute the
> LDPath with filters on them, you would get only the entities satisfying
> the condition given with the filter, wouldn't you? If so, I didn't get
> how we would use filters to partition enhancements into different
> graphs.
You are right. I forgot that the configured LDPath program is executed
several times on different Contexts within the indexed ContentItem. My
comment was in the assumption that the URI of the ContentItem is used
to apply the Filter - to decide if a ContentItem should be considered
for a semantic index or not.
> I think we still need a target graph URI. What comes to my mind
> for the trunk version is basically to pass a graph URI together with a
> chain name while sending a ContentItem to Contenthub.
>
+1
> For the branch version since users do not directly send the ContentItems
> to the indexes directly, we can pack the graph uri regarding to a
> ContentItem in a separate content part and process the graph URI within
> the Clerezza based semantic indexes accordingly so that we can partition
> the enhancements into different graphs.
Assuming that the Contenthub only supports a single Store I would
suggest that the Contenthub just adds an additional triple to the
metadata of the ContentItem
{content-item-uri} contenthub:collection {collection-uri}
and to allow using this information to filter content items during
semantic indexing. An own ContentPart seams to be an overkill.
In the long term I would like the Contenthub to support multiple
Stores and SemanticIndexes to support multiple Stores as Indexing
Sources. When adapting the Store/SemanticIndex architecture for the
Entityhub I will definitely need this. So when adding this for the
Entityhub we can than also use it for the Contenthub.
best
Rupert
--
| Rupert Westenthaler rupert.westenthaler@gmail.com
| Bodenlehenstraße 11 ++43-699-11108907
| A-5500 Bischofshofen
Re: chains/enhancers configuration
Posted by Suat Gonul <su...@gmail.com>.
Hi Rupert,
LDPath filters are not considered in the Contenthub for the time being.
As far as I understand, when you have a set of entities, and execute the
LDPath with filters on them, you would get only the entities satisfying
the condition given with the filter, wouldn't you? If so, I didn't get
how we would use filters to partition enhancements into different
graphs. I think we still need a target graph URI. What comes to my mind
for the trunk version is basically to pass a graph URI together with a
chain name while sending a ContentItem to Contenthub.
For the branch version since users do not directly send the ContentItems
to the indexes directly, we can pack the graph uri regarding to a
ContentItem in a separate content part and process the graph URI within
the Clerezza based semantic indexes accordingly so that we can partition
the enhancements into different graphs.
WDYT?
Best,
Suat
On 2/19/2013 2:24 PM, Rupert Westenthaler wrote:
> Hi Luigi
>
> On Tue, Feb 19, 2013 at 11:23 AM, Luigi Selmi <lu...@gmail.com> wrote:
>> Hi Fabian,
>>
>> as I wrote, currently it seems that all the enhancements are stored in one
>> triple collection that is configured somewhere in the code, I have still to
>> find out where exactly. The use case is simply to being able to use a
>> different triple collection/graph so that all the enhancements extracted by
>> a chain will be stored in that triple collection. This can be achieved in
>> two ways
>>
>> - as an additional parameter when a document is sent to a contenthub so
>> that the user could choose the chain the document must go through and the
>> graph in which he/she wants to store the enhancements
>> - as a configuration property of chains to be set from the Felix console
>>
>> In the second way, maybe the easiest one, it would mean to add a property
>> to enhancement engine or chain to configure the URI of the triple
>> collection to use. I have to figure out to do that at run time from the
>> Felix console but it should be possible as components that are
>> EnhancementEngine, that have a configuration policy "optional" already have
>> properties like name and ranking. Chains have configuration policy
>> "require" so maybe it would be better to add to them such property.
> Can someone tell if the Contenthub supports LDPath Filters [1]. If
> this is the case than implementing the first option should be easy.
> Just define a parameter users can specify in the request (similar to
> Title) and that is accessible by the LDPath FIlter statement.
>
> I do not like the 2nd option, because it introduces a dependency
> between the Enhancer and the Contenthub.
>
> This needs also to be seen in the context of the "new" Contenthub (see
> /STANBOL-471).
>
> best
> Rupert
>
>
> [1] http://code.google.com/p/ldpath/wiki/PathLanguage#Filter_Definition_%28Search_Indexing_only%29
>
>> Best
>>
>> Luigi
>>
>>
>> 2013/2/19 Fabian Christ <ch...@googlemail.com>
>>
>>> Hi,
>>>
>>> could you describe your use case? When would such a property makes
>>> sense and what is the desired behavior that you have in mind?
>>>
>>> Best,
>>> - Fabian
>>>
>>> 2013/2/19 Luigi Selmi <lu...@gmail.com>:
>>>> Hello,
>>>>
>>>> it seems that all enhancements are stored in a single triple collection
>>>> org.apache.contenthub.enhancements. It will be useful to add a property
>>> to
>>>> chains/enhancers to be set at run time from the felix Console to address
>>>> different triple collections. Does it make sense ?
>>>>
>>>> Best
>>>>
>>>> Luigi
>>>
>>>
>>> --
>>> Fabian
>>> http://twitter.com/fctwitt
>>>
>
>
> --
> | Rupert Westenthaler rupert.westenthaler@gmail.com
> | Bodenlehenstraße 11 ++43-699-11108907
> | A-5500 Bischofshofen
>
Re: chains/enhancers configuration
Posted by Rupert Westenthaler <ru...@gmail.com>.
Hi Luigi
On Tue, Feb 19, 2013 at 11:23 AM, Luigi Selmi <lu...@gmail.com> wrote:
> Hi Fabian,
>
> as I wrote, currently it seems that all the enhancements are stored in one
> triple collection that is configured somewhere in the code, I have still to
> find out where exactly. The use case is simply to being able to use a
> different triple collection/graph so that all the enhancements extracted by
> a chain will be stored in that triple collection. This can be achieved in
> two ways
>
> - as an additional parameter when a document is sent to a contenthub so
> that the user could choose the chain the document must go through and the
> graph in which he/she wants to store the enhancements
> - as a configuration property of chains to be set from the Felix console
>
> In the second way, maybe the easiest one, it would mean to add a property
> to enhancement engine or chain to configure the URI of the triple
> collection to use. I have to figure out to do that at run time from the
> Felix console but it should be possible as components that are
> EnhancementEngine, that have a configuration policy "optional" already have
> properties like name and ranking. Chains have configuration policy
> "require" so maybe it would be better to add to them such property.
Can someone tell if the Contenthub supports LDPath Filters [1]. If
this is the case than implementing the first option should be easy.
Just define a parameter users can specify in the request (similar to
Title) and that is accessible by the LDPath FIlter statement.
I do not like the 2nd option, because it introduces a dependency
between the Enhancer and the Contenthub.
This needs also to be seen in the context of the "new" Contenthub (see
/STANBOL-471).
best
Rupert
[1] http://code.google.com/p/ldpath/wiki/PathLanguage#Filter_Definition_%28Search_Indexing_only%29
>
> Best
>
> Luigi
>
>
> 2013/2/19 Fabian Christ <ch...@googlemail.com>
>
>> Hi,
>>
>> could you describe your use case? When would such a property makes
>> sense and what is the desired behavior that you have in mind?
>>
>> Best,
>> - Fabian
>>
>> 2013/2/19 Luigi Selmi <lu...@gmail.com>:
>> > Hello,
>> >
>> > it seems that all enhancements are stored in a single triple collection
>> > org.apache.contenthub.enhancements. It will be useful to add a property
>> to
>> > chains/enhancers to be set at run time from the felix Console to address
>> > different triple collections. Does it make sense ?
>> >
>> > Best
>> >
>> > Luigi
>>
>>
>>
>> --
>> Fabian
>> http://twitter.com/fctwitt
>>
--
| Rupert Westenthaler rupert.westenthaler@gmail.com
| Bodenlehenstraße 11 ++43-699-11108907
| A-5500 Bischofshofen
Re: chains/enhancers configuration
Posted by Luigi Selmi <lu...@gmail.com>.
Hi Fabian,
as I wrote, currently it seems that all the enhancements are stored in one
triple collection that is configured somewhere in the code, I have still to
find out where exactly. The use case is simply to being able to use a
different triple collection/graph so that all the enhancements extracted by
a chain will be stored in that triple collection. This can be achieved in
two ways
- as an additional parameter when a document is sent to a contenthub so
that the user could choose the chain the document must go through and the
graph in which he/she wants to store the enhancements
- as a configuration property of chains to be set from the Felix console
In the second way, maybe the easiest one, it would mean to add a property
to enhancement engine or chain to configure the URI of the triple
collection to use. I have to figure out to do that at run time from the
Felix console but it should be possible as components that are
EnhancementEngine, that have a configuration policy "optional" already have
properties like name and ranking. Chains have configuration policy
"require" so maybe it would be better to add to them such property.
Best
Luigi
2013/2/19 Fabian Christ <ch...@googlemail.com>
> Hi,
>
> could you describe your use case? When would such a property makes
> sense and what is the desired behavior that you have in mind?
>
> Best,
> - Fabian
>
> 2013/2/19 Luigi Selmi <lu...@gmail.com>:
> > Hello,
> >
> > it seems that all enhancements are stored in a single triple collection
> > org.apache.contenthub.enhancements. It will be useful to add a property
> to
> > chains/enhancers to be set at run time from the felix Console to address
> > different triple collections. Does it make sense ?
> >
> > Best
> >
> > Luigi
>
>
>
> --
> Fabian
> http://twitter.com/fctwitt
>
Re: chains/enhancers configuration
Posted by Fabian Christ <ch...@googlemail.com>.
Hi,
could you describe your use case? When would such a property makes
sense and what is the desired behavior that you have in mind?
Best,
- Fabian
2013/2/19 Luigi Selmi <lu...@gmail.com>:
> Hello,
>
> it seems that all enhancements are stored in a single triple collection
> org.apache.contenthub.enhancements. It will be useful to add a property to
> chains/enhancers to be set at run time from the felix Console to address
> different triple collections. Does it make sense ?
>
> Best
>
> Luigi
--
Fabian
http://twitter.com/fctwitt