IBM InfoSphere Global Name Recognition, Version 4.2 Fix Pack 1 release notes

Contents

• System requirements
• Performance considerations
• Installing Version 4.2 Fix Pack 1
• List of fixes and corrections
• Product documentation
• Known issues and changes to using the product
• Announcements

System requirements

For the latest information about hardware and software compatibility, see the detailed system requirements document at http://www-01.ibm.com/support/docview.wss?&uid=swg27019150.

Performance considerations

Searches should not exceed five tokens when running a search against a large data list (5 million or more names) on a computer that runs Linux for S/390. Including more than five tokens in your search query can return results that exceed the limit size, causing a transaction timeout that leads web services to fail. See the product information center for more performance information.

Installing Version 4.2 Fix Pack 1

Install IBM InfoSphere Global Name Recognition, Version 4.2 before attempting to install Fix Pack 1.

The installation program now writes new configuration files to *.config.template and does not overwrite existing configuration files.

Install InfoSphere™ Global Name Recognition, Version 4.2 before attempting to install Fix Pack 1.

1. Obtain the product media for the fix pack.
2. Run the installation program:

   Installation mode	Procedure
   GUI mode	Navigate to the /Disk1/InstData/VM/ directory on the product media for your platform. Run the install executable.
   Command line mode	Open a command prompt or shell window. Navigate to the /Disk1/InstData/VM/ directory on the product media for your platform. Run the install executable with the -i console option. install -i console

3. Follow the instructions on the installation program wizard to install the fix pack.
   1. On the Introduction panel, review the information and ensure that your computers meet all system requirements.
   2. On the Destination panel, browse to the directory (fully qualified path) where you want to install the fix pack. This directory must be the directory that contains your existing IBM®InfoSphere Global Name Recognition Version 4.2 installation.
   3. On the Pre-Installation Summary panel, review the summary. When you are finished, click Next to complete the installation.
   4. On the Install Complete panel, review the status, and then click Done to exit the installation.
4. If you have selected to install web services, the Name Analyzer tool, or the documentation, run the following script to start the embedded WebSphere® Application Server:

   Operating system	Script location
   Microsoft Windows	\product_install_location\startGNR.bat
   AIX®	/product_install_location/startGNR.sh
   Linux, Solaris	/product_install_location/startGNR.sh

   Attention: If the startGNR script does not exist in your installation directory, check for errors in the installation and configuration log files, which exist in the following directory:/product_install_location/installer/logs/.

List of fixes and corrections

The following list describes the errors or problems corrected in Version 4.2 Fix Pack 1:

Distributed Search

Name Preprocessor did not allow for customization of external files for parsing. No error message appears when Name Preprocessor cannot write to a file where there is not enough space.

Linguistic precision enhancements

A number of improvements have been made to name regularization, variants, and qualifiers, including:

• The generic regularization rules file, genericOnRegRule.ibm, has been updated to handle radio station frequency identifier information (for example Fiesta Radio 98.6 F.M.).
• Tokens in organizational names are now correctly regularized.
• Regularization rules changes were made for German. German spelling variations were added to the variant file.
• Enhancements to titles, affixes, and qualifiers (TAQ) include additions for Malay names.
• "M" is no longer a variant for Mohamed.

NameHunter configuration

The left bias parameter for Indian given names is now set to FALSE by default.

NameParser

Affix rules for phrase creation that were applied too strictly have been adjusted (for example, "Abdul Hussain" in India).

NameSifter

NameSifter can generate unpredictable results when used with non-English organization names. This limitation includes English names with European business suffixes such as E.U. and S.A. Do not attempt to use NameSifter with name data that is not based on United States English. It is intended for use with personal and organizational names from the United States only.

Name Preprocessor

Name Preprocessor output and configuration parameters have changed.

• Name Preprocessor no longer generates .nh files, but instead splits the interim .npp file into multiple files. As a result, the nhFile parameter and the numNhFiles parameter are replaced by the nppOutFile parameter and the numNppOutFiles parameter.
• The createNh parameter and the createNpp parameter have been removed.
  Note: Distributed Search and NameWorks Embedded Search users should re-run Name Preprocessor to generate new data list files.
• NameWorks Embedded Search users should set createOrig=false in the npp.config file to run Name Preprocessor to generate output (used as data list files).
• Distributed Search users should continue to set createOrig=true or createOrig=false, depending on whether using unique search or full search.

Search

Improved search handling of compressed names.

Transliteration

Transliteration of Japanese Kanji names is improved, including names containing spaces.

Product documentation

Known issues and changes to using the product

As problems are discovered and resolved, the IBM Support team updates the Support portal. By searching the Support portal, you can quickly find workarounds or solutions to problems.

At time of publication, there were no known installation problems. Check the Support portal for the most current information.

Memory consumption notes for Embedded Search on AIX

Memory consumption can be high on AIX machines when running Embedded Search. Possible reasons and corrective actions:

• By design, freed memory is not returned to the system on AIX. Memory is returned to the process and used as cache. As a result, memory page size can increase. Disclaiming the memory space can force memory to return to the system, but this can impact performance. It is recommended by AIX to only selectively and programmatically disclaim large allocated spaces if needed.
• Class object size tends to take more space on AIX than on other platforms. For example, a NameWorks SearchMatch object on 64 bit Linux is 72 bytes but 192 on AIX.
• A 16-byte memory alignment is used on VMX-enabled machines such as Power 6 or Power 7. This alignment can cause large memory footprint for Embedded Search which constantly allocates and uses a lot of memory.
• Bad data. Sudden memory size jumps can result from matching on empty query names. While NameHunter allows matching on empty names, the numbers of search results can be very large.
• By design, NameHunter uses memory to gain speed. Some temporary storage used in search is kept in the memory and reused. Consider the following to improve performance:
  1. Strip out leading and trailing spaces for each query name before passing it to NameHunter for search. Avoid search on empty query names.
  2. Set the environment variable, LIBCPP_NOVMX=1 on a VMX enabled machine (Power 6 or Power7) to eliminate the 16-byte padding. Note: this is only recommended for Embedded Search and the amount of memory saved is typically small.

Announcements