Pierfranco,
I have to refer this question to others here at FCLA. My high-level
understanding is that both systems (Mango and Primo - since PCI is
actually running on a copy of Primo) have their own ranking methodology
and we've attempted to normalize them to produce the results we get.
I'm out of the office right now but I've asked someone back at the
office to respond to this.
- Michele
On 9/1/2011 3:39 AM, Pierfranco Minsenti wrote:
> Hi Michele,
>
> since you have merged results in a single list in Mango (I mean: if in your
> search interface you use the first tab: All results), could you please
> explain the general methodology and decisions you have taken about the
> relevance ranking of records from different systems? I would be interested
> in knowing whether information such as the source system, the type of
> resource (e.g. a book or an article) or availability (of a book) have some
> influence on ranking.
> Thanks
> Pierfranco
>
> 2011/8/31 Michele Newberry<fclmin_at_ufl.edu>
>
>> Or you could do what we've done using the PCI API to retrieve
>> search-specific results and merge them with catalog search results. See
>> this example:
>> http://fgcu.catalog.fcla.edu/**gc.jsp?st=everglades+research&**
>> ix=kw&S=2461314820272219&fl=ba<http://fgcu.catalog.fcla.edu/gc.jsp?st=everglades+research&ix=kw&S=2461314820272219&fl=ba>
>> The non-article entries are from our Solr-based repository of records from
>> the Aleph catalog, DigiTool and other sources. The articles are from PCI
>> (without using Primo or MetaLib).
>>
>> - Michele
>>
>>
>> On 8/31/2011 3:11 PM, Molly Pickral wrote:
>>
>>> Jonathan's blog post refers to work at the University of Virginia
>>> which was, at that time, in progress. It is now in production
>>> (side-by-side search results) at:
>>>
>>> http://search.lib.virginia.edu
>>>
>>> Molly
>>>
>>> On Wed, Aug 31, 2011 at 3:07 PM, Jonathan Rochkind<rochkind_at_jhu.edu>
>>> wrote:
>>>
>>>
>>>> On 8/31/2011 2:58 PM, Paoshan Yue wrote:
>>>>
>>>>
>>>>> So does this mean that, for a library without Primo to take advantage of
>>>>> Primo Central Index, the library will need to pre-harvest content
>>>>> (catalog,
>>>>> digital collections, IR, etc.) and merge results with those from Primo
>>>>> Central? Doesn't it seem a lot of work?
>>>>>
>>>>>
>>>> A probably infeasible amount of work with less than satisfactory
>>>> results--merging results from two different search engines at
>>>> display-time
>>>> is basically the traditional "broadcast federated search", and generally
>>>> works not well at all. (And if you do want to do this, you could
>>>> alternately
>>>> buy Metalib, which comes with access to PrimoCentral-without-Primo, and
>>>> is a
>>>> broadcast federated search tool).
>>>>
>>>> But there are ways to use an aggregated index like Primo Central without
>>>> merging result sets at all.
>>>>
>>>> You could simply search PrimoCentral as it is, as an aggregated article
>>>> index, like Google Scholar or Scopus. (Perhaps it performs better in some
>>>> way than the alternatives).
>>>>
>>>> Or could provide a search which gave PrimoCentral results and results
>>>> from
>>>> one or more other search engines side-by-side ("bento box style") instead
>>>> of
>>>> merged.
>>>> http://bibwild.wordpress.com/**2011/08/08/article-search-and-**
>>>> catalog-search/<http://bibwild.wordpress.com/2011/08/08/article-search-and-catalog-search/>
>>>>
>>>>
>>>>> Cheers,
>>>>> Paoshan
>>>>>
>>>>> -----Original Message-----
>>>>>
>>>>> From: David Friggens<[log in to
>>>>> unmask]<https://listserv.nd.**edu/cgi-bin/wa?LOGON=A2%**
>>>>> 3Dind1103%26L%3DNGC4LIB%26D%**3D0%26P%3D49289&Y=yue%40UNR.**EDU<https://listserv.nd.edu/cgi-bin/wa?LOGON=A2%3Dind1103%26L%3DNGC4LIB%26D%3D0%26P%3D49289&Y=yue%40UNR.EDU>
>>>>> Subj: Re: [NGC4LIB] The next generation of discovery tools (new LJ
>>>>> article)
>>>>> Date: Mon Mar 28, 2011 6.55 pm
>>>>> Size: 1000 bytes
>>>>> To: [log in to
>>>>> unmask]<https://listserv.nd.**edu/cgi-bin/wa?LOGON=A2%**
>>>>> 3Dind1103%26L%3DNGC4LIB%26D%**3D0%26P%3D49289&Y=yue%40UNR.**EDU<https://listserv.nd.edu/cgi-bin/wa?LOGON=A2%3Dind1103%26L%3DNGC4LIB%26D%3D0%26P%3D49289&Y=yue%40UNR.EDU>
>>>>>
>>>>>> So, going back to the original topic of this thread.
>>>>>>
>>>>>>
>>>>> :-) [I've been finding the side thread on ranking to be very
>>>>> interesting and illuminating]
>>>>>
>>>>>
>>>>> To add my two cents, I've found it interesting and a little surprising
>>>>> to see that people are taking on the task of merging catalogue and
>>>>> "discovery index" data themselves.
>>>>>
>>>>> My institution/consortium chose Summon, and I've been quite glad that
>>>>> they take on that work of merging our catalogue and repository results
>>>>> along with their index. We're currently using their default interface,
>>>>> but I'm open to using another interface to display the search/results
>>>>> via the API.
>>>>>
>>>>> I can understand that if you bought Primo Central without Primo you
>>>>> would need to do result merging yourself, but at least with Summon and
>>>>> EDS I would have thought it much easier and preferable to use the
>>>>> combined results that those services provide in an alternative
>>>>> interface.
>>>>>
>>>>> Are people expecting that they can do much better by combining the
>>>>> results themselves?
>>>>>
>>>>> Cheers
>>>>> David
>>>>>
>>>>>
>> --
>> ~NOTE EMAIL ADDRESS CHANGE TO FCLMIN_at_UFL.EDU~~~~~~~~~~~~~~~~**~~~
>> Michele Newberry Assistant Director for Library Services
>> Florida Center for Library Automation 352-392-9020
>> 5830 NW 39th Avenue 352-392-9185 (fax)
>> Gainesville, FL 32606 fclmin_at_ufl.edu
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~**~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~**~~~
>>
Received on Thu Sep 01 2011 - 13:43:14 EDT