You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@stanbol.apache.org by Sawyer Chen <ch...@gmail.com> on 2013/05/17 10:38:45 UTC
How to config a Chinese enhance Chain?
Hi,
I have seen the chinese enhance chain on the demo site a few months ago and
that seems not available now. But I'm curious of how to setup a Chinese
enhance Chain.
Please point out the starting direction and/or readme/doc of creating a
chinese enhance chain which I can dig into this.
Sorry for repost if this problem has been answered before.
Sawyer
Re: How to config a Chinese enhance Chain?
Posted by harish suvarna <hs...@gmail.com>.
Thanks Rupert for all the Chinese enhancements.
FudanNLP is another LGPL software which is comparable to Stanford Chinese
package. It gives you NER and weighed keywords. Especially the chinese
names, it is very good.
Club this with dbpedia you can get some dbpedia entities. But we need
better Chinese ontologies. The chinese content in dbpedia/wiki is not as
rich compared to English.
On Fri, May 17, 2013 at 7:19 AM, Rupert Westenthaler <
rupert.westenthaler@gmail.com> wrote:
> Hi Sawyer,
>
> In this mail I will just provide an update to the original information in
> [1]
>
> * Paoding: As Paoding is not compatible to Solr 4 those
> implementations can no longer be used with the Stanbol SNAPSHOT. This
> means that you have to use Smartcn.
> * If you want to test with DBpedia you can use the index available at
>
> http://dev.iks-project.eu/downloads/stanbol-indices/dbpedia-3.8/chinese/smartcn/
> * If you what to index your own dataset with Chinese languages please
> also follow the information in [1]. But make sure to use the Smartcn
> implementations.
>
> A demo for DBpedia is available at the nightly build server at
>
> http://dev.iks-project.eu:8081/enhancer/chain/dbpedia-proper-noun
>
> As mentioned in [1] quality is currently not really good, because NLP
> is limited to sentence detection and tokenization. However I am
> currently working on the integration of Stanford NLP with Apache
> Stanbol [2]. While this integration currently only supports English I
> will extend this to also support the other languages supported by
> Stanford NLP. This also includes POS tagging and NER for Chinese. My
> current plan is to finish this work in the first or second week of
> June.
>
> With this available Linking results should be better. But still the
> Entity Linking engines are not optimized for JCK languages. Currently
> they make some assumptions that are nice for letter based languages,
> but do not necessarily work well for JCK languages.
>
> best
> Rupert
>
> [1] http://stanbol.markmail.org/thread/gqrqnl3ght2pgmob
> [2] https://github.com/westei/stanbol-stanfordnlp
>
> On Fri, May 17, 2013 at 10:38 AM, Sawyer Chen <ch...@gmail.com>
> wrote:
> > Hi,
> >
> > I have seen the chinese enhance chain on the demo site a few months ago
> and
> > that seems not available now. But I'm curious of how to setup a Chinese
> > enhance Chain.
> > Please point out the starting direction and/or readme/doc of creating a
> > chinese enhance chain which I can dig into this.
> > Sorry for repost if this problem has been answered before.
> >
> > Sawyer
>
>
>
> --
> | Rupert Westenthaler rupert.westenthaler@gmail.com
> | Bodenlehenstraße 11 ++43-699-11108907
> | A-5500 Bischofshofen
>
--
Thanks
Harish
Re: How to config a Chinese enhance Chain?
Posted by Rupert Westenthaler <ru...@gmail.com>.
Hi Sawyer,
In this mail I will just provide an update to the original information in [1]
* Paoding: As Paoding is not compatible to Solr 4 those
implementations can no longer be used with the Stanbol SNAPSHOT. This
means that you have to use Smartcn.
* If you want to test with DBpedia you can use the index available at
http://dev.iks-project.eu/downloads/stanbol-indices/dbpedia-3.8/chinese/smartcn/
* If you what to index your own dataset with Chinese languages please
also follow the information in [1]. But make sure to use the Smartcn
implementations.
A demo for DBpedia is available at the nightly build server at
http://dev.iks-project.eu:8081/enhancer/chain/dbpedia-proper-noun
As mentioned in [1] quality is currently not really good, because NLP
is limited to sentence detection and tokenization. However I am
currently working on the integration of Stanford NLP with Apache
Stanbol [2]. While this integration currently only supports English I
will extend this to also support the other languages supported by
Stanford NLP. This also includes POS tagging and NER for Chinese. My
current plan is to finish this work in the first or second week of
June.
With this available Linking results should be better. But still the
Entity Linking engines are not optimized for JCK languages. Currently
they make some assumptions that are nice for letter based languages,
but do not necessarily work well for JCK languages.
best
Rupert
[1] http://stanbol.markmail.org/thread/gqrqnl3ght2pgmob
[2] https://github.com/westei/stanbol-stanfordnlp
On Fri, May 17, 2013 at 10:38 AM, Sawyer Chen <ch...@gmail.com> wrote:
> Hi,
>
> I have seen the chinese enhance chain on the demo site a few months ago and
> that seems not available now. But I'm curious of how to setup a Chinese
> enhance Chain.
> Please point out the starting direction and/or readme/doc of creating a
> chinese enhance chain which I can dig into this.
> Sorry for repost if this problem has been answered before.
>
> Sawyer
--
| Rupert Westenthaler rupert.westenthaler@gmail.com
| Bodenlehenstraße 11 ++43-699-11108907
| A-5500 Bischofshofen