The BibRank daemon enables you to add various types of ranking based upon different ranking methods to the search engine. The ranking methods, after given the parameters necessary to find the metadata in the database, rank the documents into star categories and put them into the database. The search engine can then rank the documents it finds after receiving a query from the user, according to the choosen method(s), and show the result to the user, hopefully with more relevant documents first.
The default configurations given, are based on how the Atlantis Demo site is configured. This means that the same configuration may not apply to a normal site since it may use different tags. As an example, the 'Journal Impact Factor' method will be used. This ranking method uses the journal a document has been published in, and rank the document based on the journal's impact factor which can be accessed from various sources.
For a few of the rank methods, some external data which is not included with CDSware may be necessary to obtain. This data may for example be the 'Journal Impact Factors' from the 'Science Citation Index'. If you are using different naming convention between various sources, you can provide a convertion file so that the data can be converted to a common naming convention. For an example of a file with the necessary data for the journal impact factor method to work together with the demo installation, check out the example file 'jif.kb' below. -To create the proper files from various sources, please check out the BibRank Admin Guide and the bibrankgkb tool.
+To create the proper files from various sources, please check out the BibRank Admin Guide and the bibrankgkb tool. To be able to run a rank method, you must first add the ranking option to the correct collections, so that it appears as an option for ranking results, and add the abbreviation for the ranking method to the database. In this case we will use the abbreviation 'jif'. After adding the rank method to the database, you need to create a configuration file as shown below. (Also available below as 'jif.cfg')##First you have to add an [rank_method] category, which is mandatory for any configuration file used by bibrank [rank_method] ##Then the name of the rank method must be mentioned, only used for giving feedback when the ##program is run, it has nothing to do with the naming in the search_engine name = Journal Impact Factor ##Then the name of the ranking function needs to be given, take a look at the Bibrank Admin Guide or ##Hacking Area for ideas about the possibilities. function = single_tag_rank_method ##The size of the top star, between 0.0-1.0, the rest of the stars will split even between the range of values left. top_star_percentage = 0.10 ##In case several methods are merged into one, how important this is going to be compared to the others overall_importance = 1.0 ##Everything below depends of the function choosed above. ##For the single_tag_rank_method and authorimpact method, this category is needed [single_tag_rank] ##The name of the tag containing the needed metadata, in this case, the journal name tag = 909C4p ##The file containing, in this case, the journal name and its jif value in the form: journalname---jifvalue kb_src = /log/cdsware-DEMODEV/etc/bibrank/jif.kb ##If there is any tags that must exist for the document to be accepted by the rank method, add them here check_mandatory_tags = 909C4c,909C4v,909C4y ##If it is possible to read only modified records for the rank method. It is possible for the jif method, so yes enable_modified = yes
When the configuration file is ready and placed in the correct folder (usually etc/bibrank/), the BibRank daemon can be added to the task scheduler, or ran manually.
For the demosite, consider the following sample input data:
jif.kb -- sample data with the name of the journals and jif values.
jif.cfg -- sample configuration file, for creating the ranking stars based on journal impact factor
To create the ranking stars based on the Journal Impact Factor and add them to the database, run:
And you will be given a task number which can be used to run the method ($ bibrank 53), or you can use the various scheduling options to run it automatically. If the above configuration was used on the demosite, the statistic (if asked for in the input with -S) will look something like:$ bibrank -wjif -r Task #53 was successfully scheduled for execution.
Since this was the first run, the -r option was used, at later times, this can be avoided to speed up the execution time needed. For the creation of an up to date rankmethod without doing a full update, use '-m date', or no parameter to update records modified since last run.$ bibrank 53 2004-03-09 14:28:47 --> Task #53 started. 2004-03-09 14:28:47 --> Running: Journal Impact Factor. 2004-03-09 14:28:47 --> Statistics: Journal Impact Factor , Top Star size: 10.0% , Overall Importance: 100.0%, 2004-03-09 14:28:47 --> 0 star(s): Range>= -9.9 7990 2004-03-09 14:28:47 --> 1 star(s): Range>= -1.0 1 2004-03-09 14:28:47 --> 2 star(s): Range>= 0.964 2 2004-03-09 14:28:47 --> 3 star(s): Range>= 2.047 0 2004-03-09 14:28:47 --> 4 star(s): Range>= 3.13 2 2004-03-09 14:28:47 --> 5 star(s): Range>= 4.213 6 2004-03-09 14:28:47 --> Total: 8001
It should now be possible to rank the results after a search based on the journal impact factor.