Enterprise Portal uses an embedded search engine called the Dynamic Reasoning Engine (DRE) to perform concept-based searches. The DRE uses high-performance matching algorithms to identify patterns within text that can then be matched with similar patterns from other data sources. The identified patterns, also known as concept agents, can facilitate finding similar, relevant patterns by “learning” which search hits the user prefers.
Concept-based searching includes these features:
Natural language, Boolean, proximity, and concept queries.
Fast query speed. The speed is linked to the number of results requested rather than the number of documents indexed.
Retrieval of original document or its plain text content.
Automatic elimination of duplicates by content or reference on indexing.
You can use concept-based searching with any language to search data that has been indexed into the search engine. Several options are available to index data, which include:
WebFetch for searching Web data
Lotus Notes Fetch for searching Lotus Notes databases
ODBC Fetch for searching ODBC databases
The DRE is a scalable, multithreaded database and process based on neural-network pattern-matching technology. It performs these essential tasks:
Concept matching – accepts text as input and returns references to documents in another text source (or prebuilt index) with the most relevance.
Standard text search – accepts a Boolean term or natural language query and returns a list of documents containing the terms ordered by relevance to the query.
Stored queries – accepts a set of text (training phrase, document, or set of documents) and returns documents that match the concept. These queries can be “retrained” to make stored queries more efficient. Agents and channels are types of stored queries.
The automatic indexing utility is part of Enterprise Portal’s concept-based search service. Use the indexing utility to schedule the indexing of documents in specified data sources. You must index the data sources before you can search them.
A stored query is a search description that is stored in an Enterprise Portal database using agents and channels. An agent can be trained using natural language or by supplying documentation that describes the topic of interest. An agent can be retrained based on results, enabling future searches to be more specific and faster.
A user agent is a stored query that is created and assigned a search mission by an individual user. The user agent finds information and creates a personalized result set specifically for each user. For information about user agents, see the Portal Interface User’s Guide.
A channel is a stored query created by the system administrator. A channel is a logical category that contains subchannels. You can assign roles to a channel, which are inherited by its subchannels. For information about creating channels and subchannels, see “Performing channel, subchannel, and agent searches”.
Agents and channels are available to users with the appropriate authority, which is granted using roles. Unauthorized users are denied access to agents or channels. System administrators must create roles for an agent or channel during the same session that they create the agent or channel. If the system administrator grants permissions to a user, those permissions are not available until the user logs out and logs in again.
| Copyright © 2004. Sybase Inc. All rights reserved. |
|
|
