Tips

Decisions on when to schedule the activation of the Profiler include consideration for the following factors:

• Timeliness of data - your users will want to be notified as soon as possible after new data has been entered into the Compass Server. Try to schedule Profiler runs to run AFTER times when major Compass Server updates have taken place (that is, after robot runs bring back new data from the various web servers). For instance, you may want to schedule the first Profiler run to take place in the early hours of the morning.
• Server and network impact - The Profiler is a CPU-intensive process which could take several minutes to hours to complete (depending on the volume of new data in the database and the number of newsletters it needs to generate). It could also generate a high volume of mail traffic. As such, you should try to schedule the first run to be activated, as much as possible, during hours of minimal network traffic and server usage to minimize any potential disruptions on your network or server slowdowns. Avoid coinciding with robot runs (that is, schedule the Profiler for AFTER robot runs, not during).

We recommend that the Profiler be set to run once after the Compass Server robot is scheduled to be stopped. This is typically a large run since lots of new documents that have arrived have to be processed. You may need to allow about 2 hours for processing, so schedule the next Profiler run after that, and then as frequently as once every hour thereafter during business hours to process the feedback that is being generated.

You have the option of choosing whether to activate the Profiler on weekends. You may want to activate the Profiler during weekends if:

• The data in the Compass Server is updated during weekends and users need to know immediately
• Even if users do not need immediate updates, if there is typically high volume data updates in the Compass Server during weekends, it may be better for the Profiler to process these and generate multiple newsletters over the weekend rather than a single large one on Mondays. Note also that this is particularly important as the Profiler always returns only the first 20 matches it finds for each subscription in each users' Interest Profile.

It is easy to be confused between the two classification options. There are several very important differences between autoclassify and those of the Compass Server and you should understand what these are. The main differences are:

• in the scope of classification is that performed
• how classification rules are defined
• where the rules are stored
• when classification is performed (how it is activated).

Autoclassify classifies documents based on both the content and the header (meta-data) information of the document. It uses the full-text index created for each resource description, allowing you to specify any search query supported by the Compass Server search engine as an Autoclassify rule. As a result, you can build very sophisticated Autoclassify rules. For example, you could create a rule that said "Classify a resource description using the category "Knowledge Management Solutions:Compass server" if it contains the string "Compass" and the document does not contain references to "Needle" or "North", thus filtering out documents about a nautical compass.

Compass Server classification is performed by the robot. It also classifies documents based on its meta-tags and as well as the attributes related to content, but because it doesn't work against the full-text index, content-based classification is limited to the definition of regular string expressions.

Autoclassify rules are defined via an extended Categories Construction Editor interface that is activated either via Categories | Construction or Categories | Classification Rules Editor | Autoclassify Rules. In this extended editor, you can associate an Autoclassify rule with each category in the hierarchy (an Autoclassify rule being any valid search query supported by the Compass Server search engine).

Compass Server classification rules are defined using a separate interface called the Classification Rules Editor. You can access this interface through Categories | Classification Rules Editor | Robot Rules.

Autoclassify rules are stored as part of the categories hierarchy and are automatically deleted or updated with changes to the associated categories. Internally, these are stored in the file taxonomy.rdm located in the compass-serverid/config directory. The name of the Autoclassify attribute is "Matching-Rule". This means that as an alternative to using the enhanced Categories Editor, you can also add or edit Autoclassify rules by using a text editor and editing the taxonomy.rdm file directly. If you do this, be careful to ensure that the length of the field is entered correctly and after saving the file, restart the Compass Server (or click on Apply and then Load Configuration Files) to activate the changes.

Compass Server classification rules are stored in a separate and independent file called classification.conf. If any changes are made to the categories in the Compass hierarchy, the classification rules must be updated manually to ensure the categories they reference as in sync with those in the hierarchy.

Autoclassify performs classification at the start of each Profiler run or Autoclassify can be activated manually via Database | Autoclassify or Categories |Classify Documents | Autoclassify.

Compass Server performs classification during robot runs.

Autoclassify not only performs classifications of documents; just as importantly, it ensure that classifications it has added to the database are always in sync with categories in the category hierarchy. For example, if a category used by Autoclassify to categorize a document in the database is removed from the category hierarchy, the next activation of the Profiler (or Autoclassify) removes this category from the resource database as well. Note that Autoclassify only deletes classifications it has added, that is, invalid classifications inserted either by the robot or manually by the administrator (through the RD Editor) are not manipulated by Autoclassify.

Autoclassify works with the robot and manual classifications and will not affect classifications made through either of these avenues. However, to make the life easier for the administrator (by not having to remember to update classification rules that may either be defined by Autoclassify or Compass Server) it is generally recommended that Autoclassify be used in preference to the Compass Server's.