Adding External Metadata

External metadata Installation and configuration Juan Pablo Fernandez Ramirez.

1. Make sure you install the SQL csx_external_metadata database as per the instructions given in Installing CiteSeerX.

   Refer to the Install the SQL databases in Installing CiteSeerX

2. Download external metadata files.

   For CiteSeerX:

   a. dblp metadata: download the dblp.xml and dblp.dtd files (http://dblp.uni-trier.de/xml/). b. CiteULike: the linkouts file (http://www.citeulike.org/faq/data.adp)

3. Configure your application

   Edit conf/csx.properties to provide adequate values for:

   a. The external metadata database b. The location of the down loaded files (dblp.xml, dblp.dtd, linkouts) c. Labels: You can change the values for the labels. However, if you do so, you need also to change the values in the link_types table inside your citeseerx database to match the ones in csx.propierties file.

4. Populate/update the external database

   The database can be populated and/or updated by running:

   ./updateExternalMetadata

   The DBLP metadata is completed loaded into the database. However, from CiteULike only CiteSeerX related records are loaded.

5. How to use the external metadata database

   Currently, the external metadata is used to create links to DBLP and CiteULike from the summary pages of documents within CiteSeerX corpus. In the future, DBLP data will be used to do automatic corrections and other task that will improve CiteSeerX corpus.

   Links to external sources are stored in two tables within the CiteSeerX database:

   • links_types: which stores the label, and the base URL where the link should points to.

   • elinks: which stores the CiteSeerX paper id, the label associated to the link, and the URL portion that needs to be appended to the base URL to point to the resource in the external site.

     An example, of records in this two tables will be:

     DBLP, http://www.informatik.uni-trier.de/~ley/ and 10.1.1.1.1485 DBLP db/conf/dsn/dsn2000.html#BondavalliMCFPS00

   The elink table is populated either by:

   a. Running the ./updateExternalLinks script.

   This script will try to find matches for every document in the CiteSeerX corpus which does not have any match. It works for both CiteULike and DBLP.

   Indeed, The only way to create links for CiteULike is using this script.

   b. After a correction is done.

   The DBLP linker listen for changes in documents metadata. After a change is reported, the Linker tries to find a math between the corrected document and a record in DBLP.