You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@nutch.apache.org by Detlef Müller-Solger <d....@durato.eu> on 2008/10/08 13:23:10 UTC
Doublets
Hi,
in Germany it is reported, that one big show stopper for Nutch is the
fact, that there for example are often identical webpage’s which can be
addressed by different URLs. For example by requesting
www.xyz.de/information
or by
www.xyz.de/information/
or by
www.xyz.de/information/index
From my point of view due to the different URLs Nutch is indexing those
webpages unfortuneately three times. Is there a method to avoid the
indexing of these doublets? For example by comparing all information of
the webpage excluding the URL.
Note: A Fliter like "reduce URL generally of "/index"" is no solution
because in other cases of the same run "/index" may be needed or the
same Webpage can be adressed also by other URL Syntax.
Thanx
Detlef Müller-Solger
Re: Doublets
Posted by Julien Nioche <li...@gmail.com>.
Hi,
I haven't used it myself but it looks like the *dedup* command (
http://wiki.apache.org/nutch/bin/nutch_dedup) uses the signature of the
documents to remove duplicates. That should work fine in the case you are
describing in combination with Jasper's suggestion which would prevent
fetching some of the duplicates in the first place
Julien
--
DigitalPebble Ltd
http://www.digitalpebble.com
> On Oct 8, 2008, at 4:23 AM, Detlef Müller-Solger wrote:
>
> Hi,
>>
>> in Germany it is reported, that one big show stopper for Nutch is the
>> fact, that there for example are often identical webpage's which can be
>> addressed by different URLs. For example by requesting
>>
>> www.xyz.de/information
>> or by
>> www.xyz.de/information/
>> or by
>> www.xyz.de/information/index
>>
>> From my point of view due to the different URLs Nutch is indexing those
>> webpages unfortuneately three times. Is there a method to avoid the indexing
>> of these doublets? For example by comparing all information of the webpage
>> excluding the URL.
>>
>> Note: A Fliter like "reduce URL generally of "/index"" is no solution
>> because in other cases of the same run "/index" may be needed or the same
>> Webpage can be adressed also by other URL Syntax.
>>
>> Thanx
>>
>> Detlef Müller-Solger
>>
>>
>>
>
Re: Doublets
Posted by Jasper Kamperman <ja...@openwaternet.com>.
There is a url normalizing feature in
conf/regex-normalize.xml
For example I used the following pattern to say that pages with
&limit and &limitstart are the same on a JasperSoft forum
<!-- Jasper normalize jaspersoft URLs:
(1) &limit=6&limitstart=0 means the same as the page w/o any limit
(2) catid=10&id=NNN means the same as id=NNN&catid=10 -->
<regex-normalize>
<regex>
<pattern>(\?|\&|\&amp;)limit=6(\&|\&amp;)
limitstart=0$</pattern>
<substitution></substitution>
</regex>
<regex>
<pattern>(\?|\&|\&amp;)(id=[0-9]+)(\&|\&amp;)
(catid=10)(.*)</pattern>
<substitution>$1$4$3$2$5</substitution>
</regex>
</regex-normalize>
On Oct 8, 2008, at 4:23 AM, Detlef Müller-Solger wrote:
> Hi,
>
> in Germany it is reported, that one big show stopper for Nutch is
> the fact, that there for example are often identical webpage’s
> which can be addressed by different URLs. For example by requesting
>
> www.xyz.de/information
> or by
> www.xyz.de/information/
> or by
> www.xyz.de/information/index
>
> From my point of view due to the different URLs Nutch is indexing
> those webpages unfortuneately three times. Is there a method to
> avoid the indexing of these doublets? For example by comparing all
> information of the webpage excluding the URL.
>
> Note: A Fliter like "reduce URL generally of "/index"" is no
> solution because in other cases of the same run "/index" may be
> needed or the same Webpage can be adressed also by other URL Syntax.
>
> Thanx
>
> Detlef Müller-Solger
>
>